Nowadays, a chip multiprocessor following the x86-architecture combines at least four dedicated cores. To take benefit of this computing power, the application has to use multiple threads. To provide the parallel behavior for all processes and threads, the allocation has to change frequently. Depending on the situation, this allocation can differ and is worth to be analyzed. The application runs a configurable amount of hardworking threads. By interfering with the core-thread-allocation several different scenarios have been tested. By recording this thread-core-allocation and the execution time, it is possible to compare the different scenarios. The paper shows that Microsoft Windows 7 handles the thread-core-allocation in a lot of situations quite well. But the exclusion of core zero provides a performance increase. This is only useful if the number of threads is lower than the number of available processor cores. This situation also shows an additional interesting incident. Windows tries to balance the load over all processor cores very frequently.
|Titel||24th European Modeling and Simulation Symposium, EMSS 2012|
|Publikationsstatus||Veröffentlicht - 2012|
|Veranstaltung||The 24th European Modeling & Simulation Symposium (EMSS 2012) - Vienna, Österreich|
Dauer: 19 Sep. 2012 → 21 Sep. 2012
|Name||24th European Modeling and Simulation Symposium, EMSS 2012|
|Konferenz||The 24th European Modeling & Simulation Symposium (EMSS 2012)|
|Zeitraum||19.09.2012 → 21.09.2012|