Session Detail Information
Add this session to your itinerary

Cluster :  INFORMS Computing Society

Session Information  : Sunday Nov 07, 16:30 - 18:00

Title:  Computing with GPUs
Chair: Nick Sahinidis,John E. Swearingen Professor, Carnegie Mellon University, Department of Chemical Engineering, 5000 Forbes Avenue, Pittsburgh PA 15213, United States of America, sahinidis@cmu.edu

Abstract Details

Title: *****LATE CANCELLATION*****Solving Large-scale Dense SOCP on Heterogeneous Computing Platform
 Presenting Author: Yuriy Zinchenko,University of Calgary, Calgary, Canada, yzinchen@ucalgary.ca
 
Abstract: To minimize negative impact of uncertainties in optimal radiotherapy planning for cancer treatment, a convex robust counterpart of a conventional model has been proposed. The robust model is a large-scale dense SOCP. However, presently, such an approach is clinically infeasible due to excessive computational demands associated with solving the resulting problem. We investigate the use of heterogeneous platforms, namely GP-GPU, to speed up linear algebra operations required by IPM solver.
  
Title: Optimizing Codes for OpenCL and NVIDIA Fermi
 Presenting Author: Anne C. Elster,Norwegian University of Science and Technology/Univeristy of Texas-Austin, Norway, elster@idi.ntnu.no
 
Abstract: Modern GPUs are now massive floating-point stream processors that offer energy efficient compute power. With the recent development of tools such as CUDA and OpenCL, it has become much easier to fully utilize the computational power these system offer. This presentation will highlight some of the experiences the author’s research group has had with recent GPUs, including looking at using GPUs to compress data in order to lower latency between disk and memory and between CPU memory and GPUs.
  
Title: GPU Computing with Kaczmarz's and Other Iterative Algorithms for Linear Systems
 Presenting Author: Joseph Elble,University of Illinois at Urbana Champaign, 932 Waterview Way Apt D, Champaign IL 61822, United States of America, joseph.elble@gmail.com
 Co-Author: Nick Sahinidis,John E. Swearingen Professor, Carnegie Mellon University, Department of Chemical Engineering, 5000 Forbes Avenue, Pittsburgh PA 15213, United States of America, sahinidis@cmu.edu
 Panagiotis Vouzis,Carnegie Mellon University, 5000 Forbes Ave, Pittsburgh PA, United States of America, pvouzis@cmu.edu
 
Abstract: The GPU is used to solve large linear systems derived from partial differential equations. The PDEs studied are common to many fields, e.g. fluid dynamics and structural mechanics. The paper presents comparisons between GPU and CPU implementations of several well-known iterative methods. The results demonstrate that our GPU implementation outperforms CPU implementations of these algorithms, as well as previously studied parallel implementations on Linux clusters and shared memory systems.
  
Title: A GPU Implementation of BLAST
 Presenting Author: Panagiotis Vouzis,Carnegie Mellon University, 5000 Forbes Ave, Pittsburgh PA, United States of America, pvouzis@cmu.edu
 Co-Author: Nick Sahinidis,John E. Swearingen Professor, Carnegie Mellon University, Department of Chemical Engineering, 5000 Forbes Avenue, Pittsburgh PA 15213, United States of America, sahinidis@cmu.edu
 
Abstract: We present results with a GPU implementation of the widely used bioinformatics software BLAST. The algorithm relies on the approximate solution of dynamic programs to identify similarities in protein sequence. We present an implementation using CUDA on an NVIDIA GPU.