R. Vaswani and J. Zahorjan. The implications of cache anity on processor scheduling for multiprogrammed, shared memory multiprocessors. In Proceedings of the Thirteenth ACM Symposium on Operating System Principles, pages 26-40, December 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....such a move would be complicated, especially when the philosophers are in the midst of synchronizing. In the CP model, load balancing can happen implicitly without any change to how the processes synchronize. A potential advantage of PCM style programs is that they might bene t from cache anity [14]. That is, processes that use the same variables will be placed on the same processor, for example, as in the PCM DP example. Our work is related generally to other work that attempts to eliminate synchronization or replace synchronization by less expensive forms. Examples: eliminating barrier ....

R. Vaswani and J. Zahorjan. The implications of cache anity on processor scheduling for multiprogrammed, shared memory multiprocessors. In Proceedings of the Thirteenth ACM Symposium on Operating System Principles, pages 26-40, December 1991.

....raising request priorities [15] The CPU scheduling discipline being preemptible, there is no analog of deceptive idleness. There is, however, the equivalent of high preemption cost in switching between processes: anity scheduling attempts to schedule between manythreadsto improve cache reuse [28]. On a di erent note, non work conserving CPU schedulers have been motivated by the need to handle bursty and unexpected workloads; these are based on maintaining one or more CPUs in reserve [20] Similarly, non work conserving request schedulers have been used to support prioritized workloads in ....

R. Vaswani and J. Zahorjan. The implications of cache anity on processor scheduling for shared memory multiprocessors. In 13th ACM SOSP, Oct. 1991.

....raising request priorities [15] The CPU scheduling discipline being preemptible, there is no analog of deceptive idleness. There is, however, the equivalent of high preemption cost in switching between processes: anity scheduling attempts to schedule between many threads to improve cache reuse [28]. On a di erent note, non work conserving CPU schedulers have been motivated by the need to handle bursty and unexpected workloads; these are based on maintaining one or more CPUs in reserve [20] Similarly, non work conserving request schedulers have been used to support prioritized workloads in ....

R. Vaswani and J. Zahorjan. The implications of cache anity on processor scheduling for shared memory multiprocessors. In 13th ACM SOSP, Oct. 1991.

....that support scheduling on multiple processors will usually attempt to optimise their method of scheduling from a global ready queue. However this is certainly not optimal and still subject to system conditions at any given time. It has been shown that a global ready queue limits scalability [36]. Scalability is limited by several factors in relation to scheduling. Possibly the most signi cant in uence is that when moving a thread between processors, it is restarted with cold caches. This is important considering the ever increasing gap between memory and CPU speed. Shared data usage has ....

R. Vaswani and J. Zahorjan. The implications of cache anity on processor scheduling for multiprogrammed, shared memory multiprocessors. In Proceedings of the 13th ACM Symposium on OS Principles, pages 26-40, Pacic Grove, CA, Oct. 1991. ACM.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC