B. B. Zhou, R. P. Brent and A. Tridgell, Efficient Implementation of Sorting Algorithms on Asynchronous Distributed-Memory Machines, Report TR-CS-9306, Computer Sciences Lab, Australian National University, March 1993, 7 pp.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....model where processors do not fail, but may sometimes return the wrong result for a comparison. Related work has also been done on asynchronous computing models. Cole and Zajicek [19] proposed the APRAM model for designing parallel algorithms to work in an asynchronous setting. Zhou et al. [44] present a sorting algorithm for asynchronous machines that is not wait free. Neither is the recent sorting algorithm of Gibbons et al. 24] for the QRQW asynchronous PRAM. This sample sort based algorithm is somewhat similar to the one we use here in that it uses binary search trees built by a ....

Zhou, B. B., Brent, R. P., and Tridgell, A. Efficient implementation of sorting algorithms on asynchronous mimd machines. Tech. Rep. TR-CS-93-06, Australian National University, Computer Science Department, May 93. 19

....model where processors do not fail, but may sometimes return the wrong result for a comparison. Related work has also been done on asynchronous computing models. Cole and Zajicek [12] proposed the APRAM model for designing parallel algorithms to work in an asynchronous setting. Zhou et al. [34] present a sorting algorithm for asynchronous machines that is not wait free. Neither is the recent sorting algorithm of Gibbons et al. 15] for the QRQW asynchronous PRAM. While these models avoid making any timing assumptions, they also do not allow processor failures, and hence do not produce ....

Zhou, B. B., Brent, R. P., and Tridgell, A. Efficient implementation of sorting algorithms on asynchronous mimd machines. Tech. Rep. TR-CS-93-06, Australian National University, Computer Science Department, May 93.

....the document, the frequency with which the terms occur within the whole collection of documents and perhaps the contexts in which they occur. Methods of sorting retrieved documents have been considered in the paddy context by Bailey, Walker and Power [7] The work of Tridgell, Zhou and Brent [5] is also relevant. The ability to access the list DMA capabilities of the AP1000 hardware could speed up the sending of retrieved messages to the host. 8.4 Handling Free Form Queries Queries in the TREC workshop are expressed as topics which include a paragraph of English text describing the ....

Tridgell, A., Zhou, B. and Brent, R.P. "Efficient Implementation Of Sorting Algorithms On Asynchronous MIMD Machines," Computer Science, ANU Technical Report, TR-CS-93-06, (1993).

....a high inter processor communications latency, this algorithm could be costly, as a relatively large number of small messages are transferred. The number of messages could be reduced, but with a penalty of increased message size and algorithm complexity. One method for doing this is described in [17]. On the AP1000 the cost of the find exact algorithm was found to be very small, due to the low inter processor communications latency of the machine [7, 13] so the method of [17] was not implemented. We assume for the remainder of the discussion on the merge exchange algorithm that the ....

....but with a penalty of increased message size and algorithm complexity. One method for doing this is described in [17] On the AP1000 the cost of the find exact algorithm was found to be very small, due to the low inter processor communications latency of the machine [7, 13] so the method of [17] was not implemented. We assume for the remainder of the discussion on the merge exchange algorithm that the find exact algorithm has determined that node p must retain L 1 elements and must receive L 2 elements from node q. 3.2. Transferring Elements. After the exact number of elements to be ....

B. B. Zhou, R. P. Brent and A. Tridgell, Efficient Implementation of Sorting Algorithms on Asynchronous Distributed-Memory Machines, Report TR-CS-9306, Computer Sciences Lab, Australian National University, March 1993, 7 pp.

....proportion of the total time (see Section 3.4) On some architectures Batcher s merge exchange algorithm may not be the best choice. For example, on a systolic array it would be preferable to use an algorithm which required only nearest neighbour communication. One such variation is described in [12]. 2.5: Merge Exchange Suppose that p 1 p 2 . A merge exchange between nodes p 1 and p 2 results in node p 1 having all its elements less than or equal to those in the node p 2 , while maintaining the ordering of elements within the nodes. The efficiency of the merge exchange algorithm has a ....

....movements and 3 p N=P elements of additional storage. An important special case occurs when the sizes of the two lists are very different. Our algorithm is designed to be particularly fast in this case. Details of this algorithm are discussed in [11] and a possible improvement is given in [12]. 2 Not to be confused with Batcher s bitonic sorting algorithm [6, Sec. 5.3.4] 3: Performance 3.1: Estimating the Speedup Speedup is usually defined to be the ratio of the time taken to solve the problem on a single node (using the best known algorithm) to the time taken by the parallel ....

B. B. Zhou, R. P. Brent and A. Tridgell, Efficient Implementation of Sorting Algorithms on Asynchronous Distributed-Memory Machines, Technical Report TR-CS-93-06, CS Lab, ANU, March 1993. Available by anonymous ftp from dcssoft.anu.edu.au in the directory pub/Brent.

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