Performance estimation of streaming applications on the Intel architecture

Type: Master's Assignment
Contacts: Mark Westmijze (UT), Marc Schrijver, Rob Albers (Philips) 
Project:  
Location: CAES, University of Twente
Remark: Expired

The interventional X-Ray (iXR) developed by Philips Healthcare allow a physician to perform minimally invasive operations on a patient. Due to improved noise reduction algorithms the amount of radiation that is required during such a operation has been reduced. However, these algorithms require much processing power and the amount of processing will only increase.

 

Assignment

During the development of the new algorithms for the next generation of iXR machines it would be extremely beneficial to predict the performance of the application on new, unreleased, processors. In literature [1] a technique has been developed that can estimate the performance on a single processor. For this assignment the techniques developed by Technolution in cooperation with Philips Healthcare will be implemented to enhance the prediction model to include multi-core processing. Research question:

  • Is it possible to accurately estimate the performance of an image processing application on cache coherent multi-core systems?

 

[1] Jan Treibig, Georg Hager, and Gerhard Wellein. Complexities of performance prediction for bandwidth-limited loop kernels onmulti-core architectures. In High Performance Computing in Science and Engineering, Garching/Munich 2009, pages 3–12. 2010.

Free Joomla Templates designed by Web Hosting Top