A Tool for Binding Threads to Processors
Euro-Par '01 Proceedings of the 7th International Euro-Par Conference Manchester on Parallel Processing
Comparing the Optimal Performance of Different MIMD Multiprocessor Architectures
IPPS '98 Proceedings of the 12th. International Parallel Processing Symposium on International Parallel Processing Symposium
Bounding the minimal completion time in high-performance parallel processing
International Journal of High Performance Computing and Networking
| Hi-index | 0.00 |
The default scheduling algorithm in Solaris andother operating systems may result in frequent relocationof threads at run-time. Excessive thread relocation causepoor memory performance. This can be avoided bybinding threads to processors. However, binding threadsto processors may result in an unbalanced load.By considering a previously obtained theoreticalresult and by evaluating a set of multithreaded Solarisprograms using a multiprocessor with 8 processors, weare able to bound the maximum performance loss due tobinding threads. The theoretical result is alsorecapitulated. By evaluating another set of multithreadedprograms, we show that the gain of binding threads toprocessors may be substantial, particularly for programswith fine grained parallelism.