Dally, William J. The J-Machine: System Support for Actors. In Hewitt, Carl, and Agha Gul, editors, Concurrent Object Programming for Knowledge Pro- cessing: An Actor Perspective, MIT Press, Cambridge, MA, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is complete. This combines the simplicity of the data parallel approach with the possible performance benefits of asynchronous message passing. Concurrent Aggregates Concurrent Aggregates (CA) CD90] orignially was designed to run on Dally s fine grained message passing J machine [Dal88b] Dal88a] The intention was to place objects (called representatives) in collections (called aggregates) Concurrent messages would be fielded by the aggregate and forwarded to appropriate representatives of the aggregate. Representatives could then continue their calculations, possibly issuing ....

William Dally. The J-Machine: System Support for Actors. VLSI Memo 88-491, MIT, Cambridge, MA, 1988.

....provisions for concurrent OO style computing (Fujitsu AP1000[9] a 512 node multicomputer based on SPARC chips) we have been able to achieve an order of magnitude improvement in inter node message passing latency (approximately 35 seconds vs. a few seconds) Even compared to the Cosmos JMachine [5], which is highly optimized for concurrentOO computation, our required machine cycles are considerably smaller. Our results indicate that the concurrent object oriented computational model and languages are highly viable with a proper implementational software hardware architecture combination. ....

....saturates between 1000 to 10000 tasks, or n 15 at the above rate. 6 Discussions and Comparison to Previous Approaches Our ABCL onEM 4 software hardware architecture offers substantial performance improvements over the previous implementation architectures of OOCP languages, such as the J Machine[5] and A NET[4] Here, we would like to clarify the differences between our approach and the previous approaches, specifically how we reduce the overhead of message sends with appropriate software hardware architectural combination. In the previous approaches, the execution time was dominated by ....

William J. Dally. The J-machine: System support for Actors. In Carl Hewitt and Gul Agha, editors, Knowledge Processing: An Actor Perspective. The MIT Press, 1989.

....time for scalable execution. Since the introduction of the Actor model by Hewitt [53] in late 60 s, a number of actorbased programming systems have been developed in software on single processor or multiprocessor platforms [88, 121, 56, 13, 16, 89, 11, 70] and implemented directly on silicon [12, 37]. With the availability of low cost, high performance microprocessors, it becomes a challenge to implement actor based programming systems on stock hardware multicomputers in an efficient and scalable way [119, 117, 28, 75] It is challenging because actors are inherently concurrent and ....

....Since no state is shared among actors, it is unnecessary to provide hardware or software support for maintaining consistency of shared data. The Actor model is fairly primitive and abstract so that it may be used to model many different concurrent computing systems. For example, the J machine [36, 37] is a fine grained concurrent computer which directly implements the Actor model on its hardware. Moreover, the actor operators form a powerful set upon which to build a wide range of higher level abstractions [2] 1.2 Contributions The contributions of the thesis are summarized as follows: ffl ....

W. Dally. The J-Machine: System Support for Actors, chapter 16, pages 369--408. M.I.T. Press, Cambridge, Mass., 1990.

.... developed by Agha [2] The method has proved tractable to formal treatment and proof theory [8] and has spawned an important class of programming languages such as HAL [61] the ABCL [26, 119] series of languages and Cantor [15] and even systems support such as Rosette [111] and the J machine [40]. The actor methodology consists of independent, asynchronously communicating objects for which three basic primitives are provided: create is the equivalent of lambda abstraction in the functional paradigm but extends the dynamic resource allocation provided by function abstraction. send to is ....

W. Dally. The J-machine: System Support for Actors, In [57]. MIT Press, 1990.

....with these parameters are discussed in [Fen81] and [Fea91] 2. 1 Topology The topology we have chosen to investigate is the k ary n cube [DS87] This is a general family of graphs which has become increasingly popular among researchers and has been implemented in a number of existing machines [Dal89, Com85, Hay86, Res93]. A k ary n cube is a direct network with N = k n nodes; k is called the radix and n the dimension. A k ary 1 cube is a ring, and a k ary n cube can be built be connecting k copies of k ary (n Gamma 1) cubes in a ring. There are many variations and special cases of this network. When k = 2, for ....

W. J. Dally. The J-machine: system support for actors. In Hewitt and Agha, editors, Actors: Knowledge Based Concurrent Computing. MIT Press, 1989.

....and uses a variant of the generation scavenging algorithm. Entry tables hold interprocessor references to be used as part of the root set during GC. Each processor executes and performs GC at its own pace without interfering with the execution of other processors. The J Machine The J Machine [Dal90] is a scalable, highly parallel fine grained MIMD machine consisting of special purpose processors that are loosely coupled by means of a high speed, low latency network. The J Machine is designed to provide low overhead system services to support the actor programming paradigm [Agh86] It is ....

W. Dally. The j-machine: System support for actors. In Hewitt and Agha, editors, Towards Open Information Systems. MIT Press, 1990. To be published.

....makes little sense to put effort into generating hybrid code that would be better for the J Machine. 5. 3 Strengths and Weaknesses of the J Machine Several features of the J Machine make it excellent for running dataflow code; it was designed to support fine grained computation as described in [Dally 1988a] The features not found on most computers that proved most beneficial were: 1. Hardware support for cfutures. 2. The low latency network which gives the freedom to send frequent messages encouraging the division of tasks. 3. User defined tag types, which aided debugging. 4. The large ....

Dally, William J. The J-Machine: System Support for Actors. In Hewitt, Carl, and Agha Gul, editors, Concurrent Object Programming for Knowledge Pro- cessing: An Actor Perspective, MIT Press, Cambridge, MA, 1989.

No context found.

Dally, William J. The J-Machine: System Support for Actors. In Hewitt, Carl, and Agha Gul, editors, Concurrent Object Programming for Knowledge Processing: An Ac- tor Perspective, MIT Press, Cambridge, MA, 1989.

.... (div n 2) rb; in array (1,n) of [j] CCPgenerator radicals j) j 1 to n bound genb ; def paraffinsuntil n rb genb = radicals = radicalgenerator (div n 2) rb; in array (1,n) of [j] BCPgenerator radicals j) CCPgenerator radicals j) j 1 to n bound genb ; [0 1 0 1 0 3 0 10 0 36] def testBCPuntil n rb genb = result = BCPuntil n rb genb; in array (1,n) of [i] bv = result[i] in (length bv) i 1 to n bound genb ; 1 0 1 1 3 2 9 8 35 39] def testCCPuntil n rb genb = result = CCPuntil n rb genb; in array (1,n) of [i] cv = result[i] ....

Dally, William J. The J-Machine: System Support for Actors. In Hewitt, Carl, and Agha Gul, editors, Concurrent Object Programming for Knowledge Processing: An Actor Perspective, MIT Press, Cambridge, MA, 1989.

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