CPL - Chalmers Publication Library
| Utbildning | Forskning | Styrkeområden | Om Chalmers | In English In English Ej inloggad.

Efficient Self-tuning Spin-locks Using Competitive Analysis

Phuong Ha (Institutionen för data- och informationsteknik, Datavetenskap, Datakommunikation och distribuerade system (Chalmers)) ; Marina Papatriantafilou (Institutionen för data- och informationsteknik, Datavetenskap, Datakommunikation och distribuerade system (Chalmers)) ; Philippas Tsigas (Institutionen för data- och informationsteknik, Datavetenskap, Datakommunikation och distribuerade system (Chalmers))
Journal of Systems and Software, Elsevier Publications ( ISSN: 0164-1212). (2007)
[Artikel, refereegranskad vetenskaplig]

Reactive spin-lock algorithms that can automatically adapt to contention variation on the lock have received great attention in the eld of multiprocessor synchronization since they can help applications achieve good performance in all possible contention conditions. However, in existing reactive spin-locks the reaction relies on (i) some fixed experimentally tuned thresholds, which may get frequently inappropriate in dynamic environments like multiprogramming/multiprocessor systems, or (ii) known probability distributions of inputs. This paper presents a new reactive spin-lock algorithm that is completely selftuning, which means no experimentally tuned parameter nor probability distribution of inputs are needed. The new spin-lock is based on both synchronization structures of applications and a competitive online algorithm. Our experiments, which use the Spark98 kernels and the SPLASH-2 applications as application benchmarks, on a multiprocessor machine SGI Origin2000 and an Intel Xeon workstation have showed that the new self-tuning spin-lock performs as well as the best of hand-tuning spin-lock representatives in a wide range of contention levels.

Nyckelord: Reactive synchronization, spin-locks, shared memory multiprocessors, concurrent, performance, online algorithms.


(accepted for publication December 2006, to appear 2007)



Denna post skapades 2007-01-10. Senast ändrad 2009-12-30.
CPL Pubid: 25155

 

Institutioner (Chalmers)

Institutionen för data- och informationsteknik, Datavetenskap, Datakommunikation och distribuerade system (Chalmers) (2005-2007)

Ämnesområden

Datavetenskap (datalogi)

Chalmers infrastruktur