W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Proceedings of Supercomputing'95, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of function calls. Techniques have been proposed that use the results from type inference [47] to replace local message sends with function calls. These techniques utilize type inference to determine which methods to invoke and a locality check to ensure that the target of the message is local [33]. Unfortunately, none of these approaches have been designed to handle dynamic operations such as installation of a communicator. Such approaches derive and exploit type information during compilation whereas our approach requires run time information. However, adapting these techniques to work ....

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95, 1995. (to appear).

....ICC and the Concert system [17, 53] support grainsize optimizations through compiler analysis: the compiler increases the grainsize of objects by inlining method invocations statically, or by cloning methods and generating hints to the runtime via annotations. The actor based language HAL [54, 55] optimizes request reply communication by transforming them into asynchronous sends and extracting continuations, converts local messaging into function calls, and optimizes remote actor creation using alias addresses for actors. Mentat [17] supports dynamic placement through round robin, random ....

W. Kim and G. Agha. Efficient support of location transparency in concurrent objectoriented programming languages. In Proceedings of Supercomputing 1995, December 1995.

....global object space p p p p Thread scheduling for load balance (x 3.5) non binding concurrency p p Table 3. Runtime optimizations contributing to good parallel performance. Actor based languages [1, 20, 42, 29] are most similar in terms of high level programming support, but have focused less [39, 27] on efficient implementation. Task parallel object oriented languages, mostly based on C extensions [16, 23, 6] support irregular parallelism and some location independence, but require programmer management of concurrency, storage management, and task granularity which limits scalability and ....

W. Y. Kim and G. Agha. Efficient support for location transparency in concurrent object-oriented programming languages. In Proceedings of the Supercomputing '95 Conference, San Diego, CA, December 1995.

No context found.

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95. IEEE, 1995.

....Lisp like one to Modula like one. Finally, we implemented a runtime system with migration capability. It was initially operational on a network of DEC workstations which was connected by Ethernet and ran ULTRIX V4.2A. Then, the runtime system was ported to the Thinking Machine Corporation CM 5 [79]. During the initial period of experimentation on a CM 5, we found that the versatility of HAL caused an intolerable amount of performance overhead on program execution. The finding led us to drop support for reflection and inheritance. At this time we changed the name to THAL [103] 2 an ....

....message and attempts to consume them. When a message is dequeued for processing, the synchronization constraints of the specified method are evaluated. A positive evaluation result means the method is disabled under the actor s current state and the message is put into the actor s pending queue [79]. Recall that synchronization constraints are a function of actor state and message arguments. When an actor changes its state, some messages in its pending queue may become enabled for processing. Before yielding control over computation resources, an actor re evaluates synchronization ....

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Proceedings of Supercomputing '95, 1995.

....intra method execution computation, compile time optimizations concentrate on improving concurrent inter method execution. In particular, it restores profitable concurrency that may be lost by language constructs, such a RPC sends [2] The rest of efficiency comes from the runtime kernel [16], which provides a fast communication layer implemented using the CMAM messaging layer [26] Messages to local actors are scheduled in a stack based fashion, minimizing scheduling overhead and context switch overhead. Actor mail addresses are implemented using real addresses in a location ....

....using real addresses in a location independent way, guaranteeing locationtransparent execution of application programs. Despite the flexibility and support for dynamic computations that THAL provides, its performance on dense, regular problems is competitive with more restrictive, static languages [16]. The fine grained specification of the parallel algorithms allows us to rearrange the objects on the given processors in different ways therby composing the same parallel algorithm with different PDS s and obtaining different performance [23] The PDS s themselves are designed for specific data ....

[Article contains additional citation context not shown here]

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95, 1995. (to appear).

....the compile time optimizations of the THAL compiler concentrate on improving concurrent inter method execution. In particular, the THAL compiler restores profitable concurrency that may be lost by some high level language constructs [2] Additional efficiency is supported by the runtime kernel [9] which provides a fast communication layer implemented using the CMAM [15] Despite the flexibility and support for dynamic computations, its performance on dense, regular problems is competitive with more restrictive, static languages [9] 3 Methodology Fine grained specification of parallel ....

.... Additional efficiency is supported by the runtime kernel [9] which provides a fast communication layer implemented using the CMAM [15] Despite the flexibility and support for dynamic computations, its performance on dense, regular problems is competitive with more restrictive, static languages [9]. 3 Methodology Fine grained specification of parallel algorithms provides maximal flexibility in distributingthe workload. THAL programs may exploit this flexibility to compose a parallel algorithm with different PDS s [14] The PDS s themselves are designed for specific data structures or ....

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95, 1995.

....of non blocking asynchronous communication plus minimum context switch cost. For a set of mutually data independent request message sendings, it transforms each request into a non blocking asynchronous send and separates out their join continuation [3, 11] into a join continuation closure (JCC) [10]. Join continuation is a continuation which is dependent simultaneously on multiple replies; it is executed as all the replies are available. JCC encapsulates a join continuation and caches the context for its execution. The compiler minimizes the context to be cached through a separate data flow ....

....using two primitives, request and reply. request invokes a method on a remote actor and reply sends a result back to the requester. Figure 1 shows a Thal version of an N queen problem [13] which computes the number of the solutions. represents request message send. The Thal runtime system [10] is currently running on the TMC CM 5 [14] behv NQueen method comp(col,diag1,diag2,maxcol,depth) i, c, c1, sols, replies, where replies = zeros(N) c = col diag1) diag2) c1 = c 1) c) sols = 0; for i = 1, N do if (c1 maxcol) then break; end if ( col c1) ....

[Article contains additional citation context not shown here]

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95, 1995.

....on a single node, a naive implementation of the model would result in additional 200 overhead for each message received and a 100 overhead for each message sent. To eliminate this cost, we build on previous work aimed at optimizing concurrent object oriented languages for parallel systems [KA95, CKP93, PZC95] Specifically, we concentrate on converting communication between a meta actor and its base actor into a series of function calls. Many techniques have been proposed for using the information resulting from typeinference [PS91] to replace local message sends with function calls. ....

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Supercomputing '95, 1995. (to appear).

No context found.

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Proceedings of Supercomputing'95, 1995.

No context found.

W. Kim and G. Agha. Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages. In Proceedings of Supercomputing'95, 1995.

No context found.

Kim, W.Y. and Agha, G., "Efficient Support of Location Transparency in Concurrent Object-Oriented Programming Languages", Supercomputing'95, San Diego, USA, December, 1995.

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