
Master's Thesis
The multi-core architectures available today provide true parallel processing power in the end user domain. To leverage the constantly increasing number of cores the software industry is forced to use parallelisation techniques. One such parallelisation technique is thread-level speculation (TLS) where a thread can employ one or more helper threads to speculatively pre-execute future code sections while itself proceeds through the code at the current position. Similar to pipelining on processor level the speculatively pre-processed results may be used by the main-thread if branch-prediction was successful and the processing duration is effectively reduced. Core task of this master thesis is to develop enhancements to the Linux kernel which support thread-level speculation on application level on multi-core machines with minimal changes to the current process model.
Start date | 22. July 2012 |
End date | 16. April 2013 |
Documents | Flyer |
Projects | P'MCA |
Supervisor |
Dr. rer. nat. Holger Machens
|