You are here

Booth Talk Schedule

SC13 TOP News Release Presentations Booth Talk Schedule Booth Posters Booth Map


Tuesday, Nov. 19
11:00 ~ 11:30 GPU-acceleration of protein-protein docking and metagenome analysis on TSUBAME2.5
  Shuji Suzuki - Graduate School of Information Science and Engineering, Tokyo Institute of Technology
   
13:00 ~ 13:30 What architectures do we need for data intensive computing?
  Rick L. Stevens - Argonne National Laboratory
   
14:00 ~ 14:30 Porting and Optimizing a Real-world CFD Application with CUDA and OpenACC
  Tetsuya Hoshino - Graduate School of Information Science and Engineering, Tokyo Institute of Technology
   
15:00 ~ 15:30 Exploring Emerging Technologies in the HPC Co-Design Space
  Jeffrey S. Vetter - Oak Ridge National Laboratory, Georgia Institute of Technology
   
Wednesday, Nov. 20
13:30 ~ 14:00 TSUBAME2.5 evolution and onwards to 3.0 towards Exascale and Extreme Big Data Convergence
  Satoshi Matsuoka - GSIC, Tokyo Institute of Technology
   
14:30 ~ 15:00 AMR based on Space-filling Curve for Stencil Applications
  Takayuki Aoki - GSIC, Tokyo Institute of Technology
   
Thursday, Nov. 21
10:30 ~ 11:00 TSUBAME-KFC : Ultra Green Supercomputing Testbed
  Toshio Endo - GSIC, Tokyo Institute of Technology
   

Talk Abstracts and Speaker Bios

 

PageTop

Title GPU-acceleration of protein-protein docking and metagenome analysis on TSUBAME2.5
Abstract TSUBAME is a supercomputing system of Tokyo Institute of Technology, which has 4224 GPU cards, and has been recently updated to TSUBAME 2.5 by replacing its GPU cards with NVIDIA Kepler 20X. The system is highly powerful but GPU-computing technique is required to make full use of the system. Thus, we have developed two GPU bioinformatics applications, MEGADOCK-GPU[1] and GHOSTM[2], to take advantage of this computing power.
MEGADOCK is a protein-protein docking system, which predicts the protein complex structure from monomeric protein structures. To accelerate this application, we developed MEGADOCK-GPU by using general purpose GPU computing techniques. We implemented a system that utilizes all CPU cores and GPUs in a compute node. As results, MEGADOCK-GPU on 12 CPU cores and 3 GPUs achieved a calculation speed that was 37.0 times faster than MEGADOCK on 1 CPU core.
Also, we have developed a fast homology search tool for metagenomic analysis. A large number of sensitive homology searches are required for mapping DNA sequence fragments to known protein sequences in public and private databases during metagenomic analysis. Therefore, we developed a new efficient homology search algorithm suitable for GPU calculations that was implemented as a GPU system, GHOSTM. The system showed sufficient search sensitivity for metagenomic analysis and achieved calculation speeds that were 130 and 406 times faster than BLAST with 1 GPU and 4 GPUs, respectively.

Reference
1. Takehiro Shimoda, Takashi Ishida, Shuji Suzuki, Masahito Ohue, Yutaka Akiyama. MEGADOCK-GPU: An accelerated protein-protein docking calculation on GPUs, Parallel and Cloud-based Bioinformatics and Biomedicine, 2013.
2. Shuji Suzuki, Takashi Ishida, Ken Kurokawa, Yutaka Akiyama, GHOSTM: A GPU-Accelerated Homology Search Tool for Metagenomics, PloS one, vol. 7, no. 5, p. e36060, 2012.

Speaker Shuji Suzuki
Affiliation Graduate School of Information Science and Engineering, Tokyo Institute of Technology
Biography Shuji Suzuki is a Ph.D. student of the Graduate School at Tokyo Institute of Technology. He received a M.Eng. in Graduate School of Information Science and Engineering from Tokyo Institute of Technology in 2012. His research interests are bioinformatics, general-purpose computing on graphics processing units (GPGPU) and high performance computing.

 

PageTop

Title What architectures do we need for data intensive computing?
Abstract
Speaker Rick L. Stevens
Affiliation Argonne National Laboratory
Biography Rick L. Stevens is Associate Laboratory Director of Computing, Environment, and Life Sciences at Argonne National Laboratory, which is the U.S. Department of Energy's (DOE's) oldest lab for science and energy research. He heads Argonne's computational genomics program and co-leads the DOE laboratories planning effort for exascale computing research. He is a professor of computer science at the University of Chicago (UChicago) and is involved in several interdisciplinary studies at the Argonne/UChicago Computation Institute and at the Argonne/UChicago Institute for Genomics and Systems Biology, where he holds senior fellow appointments.
Stevens is co-principal investigator, chief technical officer, and chief architect of the DOE Systems Biology Knowledgebase project, an emerging software and data environment designed to enable researchers to collaboratively generate, test and share new hypotheses about gene and protein functions, perform large-scale analyses on a scalable computing infrastructure, and model interactions in microbes, plants, and their communities. Stevens is also co-principle investigator for the NIAID Bioinformatics Resource Center program where his group has developed computational tools and genomics databases to support infectious disease research.
Stevens is interested in the development of innovative tools and techniques that enable computational scientists to solve important large-scale problems on advanced computers. His research focuses on two principal areas: high-performance computer architectures, and computational problems in the life sciences. In addition to his research work, Stevens teaches courses on computer architecture, collaboration technology, parallel computing, and computational science. He serves on many national and international advisory committees and still finds time to occasionally write code and play with his 3D printer.

 

PageTop

Title Porting and Optimizing a Real-world CFD Application with CUDA and OpenACC
Abstract Computational Fluid Dynamics (CFD) applications are one of the most important applications executed on supercomputers. Especially, GPU-based supercomputers have been showing remarkable performance of CFD applications. However, GPU-programing is still difficult to obtain high performance, which prevents legacy applications from being ported to GPU environments.
OpenACC is a OpenMP-like directive-based programming interface and CUDA is the most widely used programming language for accelerators. This talk presents the result of the performance evaluation of porting a Large-Scale CFD application UPACS with CUDA and OpenACC.
Speaker Tetsuya Hoshino
Affiliation Graduate School of Information Science and Engineering, Tokyo Institute of Technology
Biography Tetsuya Hoshino is a graduate student of Tokyo Institute of Technology, Tokyo, Japan, studying at Prof. Satoshi Matsuoka laboratory. He received his B.S. with an emphasis in Computational Fluid Dynamics and GPGPU in 2012 from Tokyo Institute of Technology. His current research interest is programming language for heterogeneous supercomputers that installed many-core accelerators; such as PGI Accelerator, HMPP, OpenACC, and CUDA. He is particularly interested in the portability and performance of directive-based approach and compiler-based optimizations by auto-tuning.

 

PageTop

Title Exploring Emerging Technologies in the HPC Co-Design Space
Abstract New architectures, such as novel heterogeneous cores and NV-RAM memory systems, are often radically different from today's systems. Our team has recently developed a number of techniques for modeling, simulating, and measuring these future systems. Aspen, our performance modeling language, allows users to compose and answer modeling questions quickly.
Speaker Jeffrey S. Vetter
Affiliation Oak Ridge National Laboratory, Georgia Institute of Technology
Biography Jeffrey Vetter, Ph.D., holds a joint appointment between Oak Ridge National Laboratory (ORNL) and the Georgia Institute of Technology (GT). At ORNL, Vetter is a Distinguished R&D Staff Member, and the founding group leader of the Future Technologies Group in the Computer Science and Mathematics Division. At GT, Vetter is a Joint Professor in the Computational Science and Engineering School of the College of Computing, the Principal Investigator and Project Director for the NSF Track 2D Experimental Computing Facility for large scale heterogeneous computing using graphics processors, and the Director of the NVIDIA CUDA Center of Excellence. His papers have won awards at the International Parallel and Distributed Processing Symposium and EuroPar; he was awarded the ACM Gordon Bell Prize in 2010. His website contains more information: http://ft.ornl.gov/~vetter/.

 

PageTop

Title TSUBAME2.5 evolution and onwards to 3.0 towards Exascale and Extreme Big Data Convergence
Abstract TSUBAME2.5 is an upgrade to TSUBAME2.0 commissioned in November 2010, by replacing all the 4224 NVIDIA Tesla M2050 GPUs to the latest Kepler K20x, with peak performance of 5.76 and 17.1 Petaflops in double and single point precision arithmetic respectively, latter being the fastest metric in Japan. By overcoming several technical obstacles, TSUBAME2.5 has demonstrated x2-3 speedup for many applications on the whole machine, demonstrating multi-petascale performance, as well as being a valuable stepping stone for TSUBAME3.0 in 2015-16. Another theme for both TSUBAME2 and TSUBAME3 are convergence with Extreme Big Data where we expect to realize I/O capabilities in many TeraByte/s which is on par with the bandwidth entire global Internet, as a result of our new JST-CREST project "Extreme Big Data" in which we seek convergence of supercomputing and IDC Big Data processing.
Speaker Satoshi Matsuoka
Affiliation GSIC, Tokyo Institute of Technology
Biography Satoshi Matsuoka is a Professor at the Global Scientific Information and Computing Center of Tokyo Institute of Technology (GSIC). He is the leader of TSUBAME series of supercomputers, which became the 4th fastest in the world on the Top500 and awarded the "Greenest Production Supercomputer in the World" by the Green 500 in November, 2010 and June 2011. He has also co-lead the Japanese national grid project NAREGI during 2003-2007, and is currently leading various projects such as the MEXT Green Supercomputing and JSPS Billion-Scale Supercomputer Resilience, as well as the MEXT/JST Feasibilty study for Exascale, and now the JST CREST Extreme Big Data. He has authored over 500 papers according to Google Scholar, and has chaired many ACM/IEEE conferences, including the Technical Papers Chair, Community Chair, and the upcoming Program Chair for Supercomputing Conferences 09, 11 and 13 respectively. He is a fellow of ACM and European ISC, and has won many awards including the JSPS Prize from the Japan Society for Promotion of Science in 2006, awarded by his Highness Prince Akishinomiya, the ACM Gordon Bell Prizes for 2011, and the Commendation for Science and Technology by the Minister of Education, Culture, Sports, Science and Technology in 2012. .

 

PageTop

Title AMR based on Space-filling Curve for Stencil Applications
Abstract AMR is an efficient method to assign a mesh with a proper resolution to any local areas. By using bigger leaves than those of CPU, we can assign a CUDA block to a leaf with enough thread numbers. We show a GPU implementation in which the leaves are connected by a space-filling curve.
Speaker Takayuki Aoki
Affiliation GSIC, Tokyo Institute of Technology
Biography Takayuki Aoki received a BS in Applied Physics (1983), an MSc in Energy Science and Dr. Sci. (1989) from Tokyo Institute of Technology, has been a professor in Tokyo Institute of Technology since 2001 and the deputy director of the Global Scientific Information and Computing Center since 2009. He received the Minister Award of the Ministry of Education, Culture, Sports, Science & Technology, the Computational Mechanics Achievement Award from Japan Society of Mechanical Engineers, the Achievement Award of the Japan Society for Industrial and Applied Mathematics and many awards and honors in GPU computing, scientific visualization, and others. His team got the Gordon Bell Prize in 2011. He was also recognized as a CUDA fellow by NVIDIA in 2012.

 

PageTop

Title TSUBAME-KFC : Ultra Green Supercomputing Testbed
Abstract Tokyo Tech GSIC has constructed a new supercomputing testbed, named TSUBAME-KFC, in order to promote research to make next-gen supercomputers even low-power. In this testbed, high-density GPU computing nodes are cooled by a new liquid-submersion cooling method, where the entire compute nodes are submerged in the coolant liquid. We report power efficiency and computing performance of TSUBAME-KFC.
Speaker Toshio Endo
Affiliation GSIC, Tokyo Institute of Technology
Biography Toshio Endo is an associate professor at the Global Scientific Information and Computing Center (GSIC) of Tokyo Institute of Technology. He is also a leader of JST(Japan Science and Technology Agency) CREST project, "Software Technology that Deals with Deeper Memory Hierarchy in Post-petascale Era". His research interests include high performance and low power computing. He has a Ph. D. in science from the University of Tokyo (2001).
PageTop