Scientific and technical positions in ARPA team
Master internship (2011): Study of the influence of Linux operating system on OpenMP applications performances on multicore processors
This research subject is devoted to understand the interaction between the application, the operating system and the underlying multicore processor. Our recent study on this field shows that the variability of program execution times is important, see:
- Abdelhafid Mazouz and Sid-Ahmed-Ali Touati and Denis Barthou. Measuring and Analysing the Variations of Program Execution Times on Multicore Platforms: Case Study. Research report, University of Versailles St-Quentin en Yvelines, number HAL-INRIA-00514548. July 2010. hal.archives-ouvertes.fr/inria-00514548
If the number of threads does not exceed the number of cores, then the micro-architectural effects of the processor dominate the performance of the code. However, if the number of threads exceeds the number of cores, then the interaction with the OS dominate the performance. We have observed that slight updates in the linux kernel bring many performance benefits. The subject of this master internship is to study in details what are the factors in the OS that influence the execution time variability of an OpenMP application. We will start the study by focusing on process scheduling strategies, synchronisation. The student may however need to study other OS services that explain the observed instability of program’s execution times.
The master student must be familiar with OpenMP, processor architecture and micro-architecture, operating system.
This master internship has a grant and opens opportunity to continue to PhD studies.
Contact: Sid DOT Touati AT uvsq DOT fr
Need to study in a master of HPC ?
The ARPA team is involved in the following master specialised in high performance computing.