To main content

How to program efficient optimization algorithms on Graphics Processing Units - The Vehicle Routing Problem as a case study

Abstract

Modern PCs are parallel and heterogeneous, with a growing number of cores for task parallelism and ever more powerful data parallel accelerators. The glory days when you could go to the beach for two years, and your sequential optimization program would run twice as fast on a new PC when you got back, are gone due to technological limits. To utilize the computing power of modern PCs and profit from future hardware developments you need to rewrite your sequential code and even rethink your optimization methods. I will go through the recent change in processor development and its repercussions. The potential of heterogeneous computing will be illustrated with fresh results from a GPU-based VRP solver developed at SINTEF. Speculative ideas for a self-adaptive, heterogeneous VRP solver will conclude my talk.

Category

Academic lecture

Client

  • Research Council of Norway (RCN) / 205298

Language

English

Affiliation

  • SINTEF Digital / Mathematics and Cybernetics

Presented at

Seminar (invited talk)

Place

Newcastle, Australia

Date

06.07.2011 - 06.07.2011

Organizer

University of Newcastle

Year

2011

View this publication at Cristin