This paper reports the results on comparing the asynchronous QRQW model to the bulk-synchronous QRQW model, and some open problems that remain. It also describes the features of the two models. In the bulk-synchronous QRQW PRAM, each processor executes a series of reads to shared memory locations, and the synchronizes with all the other processors. The time for such a bulk-synchronous step is the maximum of the number of reads, compute steps, and writes by any one processor and the maximum contention at any one location. In the QRWR ASYNCHRONOUS PRAM, each processor executes reads, compute steps, and writes but processors proceed at their own pace with no intervening synchronization. Algorithms must be correct under worst case assumptions on the finite delays incurred by processors and in processing memory requests.
|Number of pages||1|
|State||Published - 1996|
|Event||Proceedings of the 1996 15th Annual ACM Symposium on Principles of Distributed Computing - Philadelphia, PA, USA|
Duration: 23 May 1996 → 26 May 1996
|Conference||Proceedings of the 1996 15th Annual ACM Symposium on Principles of Distributed Computing|
|City||Philadelphia, PA, USA|
|Period||23/05/96 → 26/05/96|