Title: Partitioning Spatially Located Computations using Rectangles

Speaker: Prof. Umit Catalyurek, Ohio State University

Date/Time: Thursday, October 21, 2010, 10:00 am Mountain Time       

Location: CSRI Building/Room 90 (Sandia NM)

Brief Abstract: Distributing spatially located computations on a parallel computer is an important step to efficiently use computing resources on a large class of scientific applications. In this talk we will investigate different classes of rectangle-based partitions to balance the computations among processors. I will present a quick survey of existing methods, introduce new classes of partitions, propose new optimal and heuristic algorithms, and  provide worst case performance analyses for some algorithms. I will also present experimental evaluation of the newly proposed algorithms against the existing methods.

Biography:     Umit V. Catalyurek is an Associate Professor in the Departments of Biomedical Informatics and Electrical & Computer Engineering, at the Ohio State University. Dr. Catalyurek received his Ph.D. in 2000 from Bilkent University, and prior to joining the Ohio State University in 2001, he worked at Johns Hopkins University as a Research Associate and at the University of Maryland as a Visiting Research Scientist. Dr. Catalyurek is a recipient of NSF CAREER award and primary investigator of several awards from Department of Energy, National Institute of Health, and National Science Foundation. Dr. Catalyurek is the General Chair of 1st ACM International Health Informatics Symposium 2010, and Workshops Chair of IPDPS 2010 and 2011. Dr. Catalyurek also serves in the editorial board of the journal Parallel Computing, and on the program committees and organizing committees of several international conferences. Dr. Catalyurek has co-authored over 100 peer-reviewed articles, and his main research areas are in high performance computing, combinatorial scientific computing and biomedical informatics.

CSRI POC: Erik Boman, 505-844-2003