Laxmikant V. Kal e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua M. Yelon. Converse: an interoperable framework for parallel programming. In Proceedings of the International Parallel Processing Symposium, pages 212--217, 1996. Available from http://charm.cs.uiuc.edu/papers/ converse-ipps96.ps.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....approach to writing wide area parallel (grid computing) applications. Most other grid computing systems (e.g. Globus [11] and Legion [15] support a variety of languages. GridLab [30] is building a toolkit of grid services that can be accessed from various programming languages. Converse [16] is a framework for multi lingual interoperability. The SuperWeb [1] Javelin 2.0 [24] and Bayanihan [29] are examples of global computing infrastructures that support Java. A language centric approach makes it easier to deal with heterogeneous systems, since the data types that are transferred ....

Laxmikant V. Kale, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua Yelon. Converse: An interoperable framework for parallel programming. In Intl. Parallel Processing Symposium, 1996.

....Both expose memory registration at the API level. Both have notions of ordered units of data sent within larger, unordered units. And both are connectionoriented but support fast, logical connections within an established physical connection. The ability to grab slabs comes from Converse [17], which has an analogous feature. V. CONCLUSIONS Thanks to user level messaging layers, nearhardware communication performance is often available to applications running within a SAN. We claim that availability, usability, and management are the next logical step needed to make wide area ....

Laxmikant V. Kald, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua Yelon, "Converse: An interoperable framework for parallel programming," in Proceedings of the loth International Parallel Processing Symposium (IPPS '96), Honolulu, Hawaii, Apr. 15 19, 1996, pp. 212 217, Available: http://charm. cs. uiuc. edu/paper s/Inter0plPPS96. ps.

....host. Other systems like IceT [Gray 97] and InfoSphere [Chandy 97] see the Internet as a cooperative or group environment in which multiple users can work together. The IceT project has also done some work at providing a PVM like environment for parallel processing. The parallel Java extension [Kal 97] uses the Converse [Kal 96] interoperability framework, which makes it possible to integrate parallel libraries written in Java with modules in other parallel 4 languages in a single application. The main drive behind these projects was the recognition of the write once run many capability of the ....

Kal, L. V., Bhandarkar, M., Jagathesan, N., Krishnan, S. and Yelon, J. Converse: An Interoperable Framework for Parallel Programming. Proceedings of the 10 th International Parallel Processing Symposium, Honolulu, Hawaii, April 1996, pp.212217. 36

....and library developers rather than to end users. The range of applications being considered for development on FM, or already in the works, is testimonial of the flexibilty of its interface. Besides the MPI message passing library, these are the BSD socket interface, the Converse compiler backend [6], the Orca Project parallel object language [2] LogP LogP is a model for parallel algorithm design proposed by Culler et al. in [4] It is intended to serve as a basis to develop fast, portable parallel algorithms and to offer guidelines to machine designers. The LogP model is based on four ....

Laxmikant V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua M. Yelon, "Converse: an interoperable framework for parallel programming", In Proceedings of the International Parallel Processing Symposium, 1996,

....and library developers rather than to end users. The range of applications being considered for development on FM, or already in the works, is testimonial of the flexibility of its interface. Besides the MPI message passing library, these are the BSD socket interface, the Converse compiler backend [15], the Tempest runtime library [12] the Orca Project parallel object language [3] FM performance exceeds the performance of messaging layers presently available on a number of MPPs [14, 24, 8] FM 1.1 is the version which has been used to build the MPI library upon. The two graphs of figure 2.1 ....

Laxmikant V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua M. Yelon. Converse: an interoperable framework for parallel programming. In Proceedings of the International Parallel Processing Symposium, 1996. Available from http://charm.cs.uiuc.edu/papers/converse-ipps96.ps.

....accessible from other objects. We have not addressed this problem, because we think that the internal state of an object should not be seen nor modi ed without using an object s method. This approach is followed by many object oriented languages (for example Eioeel) INRIA 11 the Converse [11] interoperability framework, which makes it possible to integrate parallel libraries written in Java with modules in other parallel languages in a single application. Libraries and frameworks : other projects use the Java language without any extension. Some environments rely on a data parallel ....

L. V. Kal#, M. Bhandarkar, N. Jagathesan, S. Krishnan, and J. Yelon. Converse : an interoperable framework for parallel programming. In Proceedings of the 10th International Parallel Processing Symposium, pages 212217, Honolulu, Hawai, April 1996.

....and library developers rather than to end users. The range of applications being considered for development on FM, or already in the works, is testimony to the flexibility of its interface. Besides the MPI message passing library, these are the BSD socket interface, the Converse compiler back end [12], the Tempest runtime library [11] the Orca Project parallel object language [2] MPI and MPICH MPI (Message Passing Interface) is a message passing library, with primitives specifications for both C and Fortran. It is the result of the work of the MPI Forum, a committee composed of vendors and ....

Laxmikant V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua M. Yelon. Converse: an interoperable framework for parallel programming. In Proceedings of the International Parallel Processing Symposium, 1996. Available from http://charm.cs.uiuc.edu/papers/converse-ipps96.ps.

....implementation and level of support as well as subjective preferences play a significant part in the selection of languages for an application. Recognizing that different 10 languages may be better suited for different modules even within a single application, runtime frameworks such as Converse [20] support multi paradigm interoperability. 4 Concluding Remarks What languages will continue to be used in future CSE applications It is clear that parallel machines will be a part of the landscape of computing in future. Automatic parallelizing compilers that can target scalable parallel ....

L. V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua Yelon. Converse: An Interoperable Framework for Parallel Programming. In Proceedings of the 10th International Parallel Processing Symposium, pages 212--217, Honolulu, Hawaii, April 1996.

....algorithm, threads have several drawbacks. Extra memory is required for multiple stacks, there is overhead from context switching between threads, and a running sequencer cannot migrate between processors. These problems will be addressed by using the Structured Dagger coordination language (Kal e and Bhandarkar, 1996) which enables programmers to specify partial order between entry methods of an object. Using constructs such as overlap, forall, and when blocks, one can easily express dependencies between entry methods of an object while letting the system do the buffering, maintaining counters, etc. required ....

Kal'e, L. V., M. Bhandarkar, N. Jagathesan, S. Krishnan, and J. Yelon. 1996. Converse: An interoperable framework for parallel programming. In Proceedings of the 10th international parallel processing symposium.

....migration alone provides a good final load balance. During the simulation, each migratable object informs the load balance coordinator when it begins and ends execution, and the coordinator accumulates the total time consumed by each migratable object. Furthermore, the Converse runtime system [6] provides callbacks from its central message driven scheduler, which allows the coordinator to compute idle time for each processor during the same period. All other computation, including the bond force computations and integration, is considered background load. The time consumed by the ....

L. V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua Yelon. Converse: An Interoperable Framework for Parallel Programming. In Proceedings of the 10th International Parallel Processing Symposium, pages 212--217, Honolulu, Hawaii, April 1996.

....ensure that each language pays for only those features that it uses, the Converse architecture consists of a set of very simple, efficient core components. Higher level functionality is layered on top of this core in the form of optional modules. The idea of need based cost is discussed further in [11]. The set of core components is chosen so that a wide variety of programming models, and their coexistence, can be supported. Converse models the machine as a set of semi independent processors that communicate primarily via messages. Each processor is running a scheduler, which is responsible for ....

L. V. Kal'e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua Yelon. Converse: An Interoperable Framework for Parallel Programming. In Proceedings of the 10th International Parallel Processing Symposium, pages 212--217, Honolulu, Hawaii, April 1996.

....a portable manner. Charm , which was developed in 1992 93 [12] fully incorporates object oriented features such as inheritance and polymorphism into the Charm model and adds other abstractions such as parallel object arrays. Recently, the Chare Kernel was modified to operate on top of Converse [13], a machine independent layer (described in section 2.3.1) which supports interoperability by allowing modules from multiple languages to co exist in a single application. 2.2 The Charm language In this section we discuss the essential features of Charm . The first two ....

....Charm has been implemented using a translator and a runtime system. The translator converts Charm constructs into C constructs and calls to the runtime system, and also generates stubs for messaging. The runtime system consists of a language independent portable layer called Converse [13], on top of which is the Chare Kernel layer [8, 11] Figure 2.3 shows the structure of the runtime system. 2.3.1 Converse: portability and interoperability The Converse layer provides a portable machine interface which supports essential parallel operations on MIMD machines. These include ....

L.V. Kal'e, M. Bhandarkar, N. Jagathesan, and S. Krishnan. Converse: An Interoperable Framework for Parallel Programming. Technical Report 95--12, Parallel Programming Laboratory, Department of Computer Science, University of Illinois, Urbana-Champaign, March 1995.

....properties of parallel computers. Converse commits to a model which acknowledges the fact that several levels of sharing are possible, each with its own advantages. Converse therefore provides primitives to declare and access data at each level of sharing. 2 Converse Threads Subsystem Converse [4] is a machine interface for parallel systems and languages. It seeks to achieve the following major goals: 1. portability: that programs written on top of Converse be executable without modification on a wide range of platforms, 2. generality: to be able to implement the full spectrum of parallel ....

....not found in other parallel runtime systems. Chief among these resources is CPU time itself. An important part of Converse s mechanisms for allocating and managing CPU time is the Converse threads subsystem, which is the focus of this paper. The other parts of Converse are discussed elsewhere [4]. The Converse threads subsystem is not a threads package in the traditional sense of the phrase. It includes some software, to be sure, but its goal is not just to provide some useful subroutines. Instead, it seeks to guarantee that software written with the Converse threads subsystem will be ....

[Article contains additional citation context not shown here]

L.V. Kal'e, M. Bhandarkar, N. Jagathesan, S. Krishnan, and J. Yelon. Converse: An Interoperable Framework for Parallel Programming. In International Parallel Processing Symposium 1996 (to appear), 1996.

No context found.

Laxmikant V. Kal e, Milind Bhandarkar, Narain Jagathesan, Sanjeev Krishnan, and Joshua M. Yelon. Converse: an interoperable framework for parallel programming. In Proceedings of the International Parallel Processing Symposium, pages 212--217, 1996. Available from http://charm.cs.uiuc.edu/papers/ converse-ipps96.ps.

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