| J.Trescher, F.Bieler, C.Hinrichs. "Modula-L: Implementation of the Linda Model for Arbitrary Transputer Networks", Parallel Computing: From Theory to sound Practice, IOS Press, 1992. |
....withdrawals to predetermined in out sets. However, it was implemented in transputer networks [13] and the scheme proved to be extremely inefficient due to high communication overhead. There is another popular solution based on hashing , which is used in the Linda kernel for the iPSC and Modula L [30]. Tuples are distributed across the network according to an hashing function applied to a tuple typesignature. This scheme has the advantage of avoiding broadcasts and reducing storage costs due to replication. However, an hashing function is valid only for a specific topology and size. If we ....
J.Trescher, F.Bieler, C.Hinrichs. "Modula-L: Implementation of the Linda Model for Arbitrary Transputer Networks", Parallel Computing: From Theory to sound Practice, IOS Press, 1992.
....differently. One exception was the Sieve of Erasthotenes, where having two search fields would have helped. With some other programs the need for a second key field is also felt. In rarer cases, a third field might be useful. An important observation in this context, also reported for instance by [30], is that the fields used to specify an associative search tend to be always the same. We have found no case where a field used to specify one associative search is a formal parameter in another associative search. Moreover, many master worker programs tend to have only two tuple types: tuples ....
....with 8 workers 24 s 305 ms 68 s Columns 2 and 3 lists figures for Modula L, a Linda programming environment having Modula 2 as a base language, that implements a distributed Tuple Space using hashing techniques. The results presented were obtained with a system quite similar to ours [30]. The best worst case column results were obtained for an optimal worst placement of tuples in the nodes. Column 4 lists figures for the S Net Linda Kernel. Finally column 5 lists figures for a MINIX Linda implementation[12] The good performance of ParLin can be explained in part due to: 1) the ....
[Article contains additional citation context not shown here]
J.Trescher, F.Bieler, C.Hinrichs. "Modula-L: Implementation of the Linda Model for Arbitrary Transputer Networks", Parallel Computing: From Theory to sound Practice, IOS Press, 1992.
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