Dennis Gannon and Jeng-Kuen Lee. Object-Oriented Parallel Programming: Experiments and Results. In Proceedings of Supercomputing '91 . IEEE, 1991.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....expensive dot products, to complex heuristics. Specifying a data parallel computation in terms of a single element is the approach we have used in creating our data parallel extensions. We call this approach elementcentered. Fundamentally, this concept is not new to data parallel languages [2, 4, 6, 13, 15, 16, 17, 20, 21]. However, we have extended the notion to encompass subset level data parallelism. By subset level data parallelism we mean allowing the definition of operations in which subsets (as opposed to elements) are the data granules, e.g. a row or a column. This is not the same as applying an ....

....mentat object member function implementation executing concurrently with other mentat object member functions. Figure 6 illustrates both task and data parallel invocations within the same code fragment. 5: Related Work In terms of purely data parallel languages, Dataparallel C [13, 20, 21] pC [2, 16], C [15] Fortran D [6] Fortran 90 [1] and High Performance Fortran (HPF) 17] are the languages from which are related to our work. C and pC are based on C . Dataparallel C is based on C, but uses some ideas from object oriented language design. HPF s origin is Fortran. Our work ....

J.K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, 1991, Albuquerque, NM, pp. 273-282.

....old code is rewritten, bugged and debugged time and again. 2.2. Collection oriented Languages There are now many languages in various stages of development which handle arrays of data more effectively, such as High Performance Fortran[6] C [14] APL[5] Paralation Lisp[1] C [9] pC [10], LPAR[8] The code a=3 b can now be interpreted as multiplying each member of the array b by 3 and setting it to the corresponding element of the array a. These languages are a great help in simplifying the writing, understanding, or debugging of software, as well as allowing the compiler and ....

J. K. Lee and D. Gannon, Object oriented parallel programming experiments and results, p 273 in Proc. Supercomputing `91, ACM Press, 1991.

.... annotation transparent with regular parallelism [8] PANDORE extra annotation transparent with regular parallelism [9] Crystal extra annotation compiler dependent [10] MENTAT mixed in a single class modification may be required [11] pC (old) mixed in a single class modification may be required [12] EPEE Paladin mixed in a single class modification may be required [13] ScaLAPACK mixed in a single class modification may be required [14] Charm mixed in a single class modification may be required [15] Ours encapsulated in the hierarchy transparent Gamma Table 2. Available systems for ....

J.K. Lee and Dennis Gannon. Object Oriented Parallel Programming Experiments and Results. In Supercomputing '91, 1991.

....(SPMD) characterizes this kind of computation more appropriately than SIMD. The traditional array based data parallel model[19] has been integrated into a number of different programming languages (e.g. FORTRAN 90[27] C [18] etc. This includes extensions to object oriented languages (pC [22]) Recognizing the performance penalty of a completely synchronous data parallel constructs on modern MIMD machines, asynchronous (SPMD) FORALL constructs have been introduced into more recent languages[41, 17] The design of class libraries is fundamental to the object oriented approach. The ....

J. K. Lee and D. Gannon. Object Oriented Parallel Programming Experiments and Results. In Proceedings of Supercomputing '91. IEEE Computer Society Press and ACM SIGARCH, November 1991.

....perturbation of the original serial code. 2.6 Other Models and Implementations In this section, we survey and briefly compare the approaches represented in this work to nonactor models and implementations. Other actor approaches are considered in Section 3.6. 2.6. 1 pC Gannon and Lee [21] [22], 23] 24] have developed pC , an extension of C with support for distributed data structures similar to FORTRAN D. The pC language provides support for distributed collections, both array based and tree based, of arbitrary types and with full support for the C mechanisms of ....

J. K. Lee and D. Gannon, "Object oriented parallel programming experiments and results," in Proceedings, Supercomputing '91, pp. 273--282, Nov. 1991.

....technology motivates implementing a COOP language by extending an existing sequential object oriented language, such as C or Smalltalk, with a notion of process or thread. In particular, given the popularity and portability of C ,a number of COOP languages based on C have proliferated [23, 69, 47, 73, 27, 87, 86, 95]. We describe a few examples below. Compositional C (CC ) 69] extends C with a number of abstractions for process creation and synchronization. Synchronization is done via special shared variables. COOL [27] is targeted for shared memory multiprocessors. Invocation of a parallel function ....

....objects from sequential ones; programmers are required to specify what classes are to be executed in parallel. Mentat objects map one to one onto processes in a virtual machine. By contrast, Charm requires programmers take the responsibility of mapping of objects onto processing nodes. pC [87], C [86] and pSather [95] are all C based COOP languages which are designed to support data parallelism. They differ in how to initiate data parallel execution. CST [61, 38] and DistributedConcurrentSmalltalk (DCS) 96] are two of many COOP languages which extended Smalltalk 80 [43] CST ....

[Article contains additional citation context not shown here]

J. K. Lee and D. Gannon. Object-Oriented Parallel Programming Experiments and Results. In Proceedings Supercomputing 91, pages 273--282, 1991.

....A Better Matrix : 30 1 Introduction 1. 1 Existing Parallel Machines and Languages There are many different parallel languages available today for a variety of parallel architectures and models; examples include Mentat[7] PC [8], Modula 3 [9] and Split C[10] Using these languages one can soon be overwhelmed by their complexity and variance: The languages are different, often vastly so. Few are designed to conform to a model such as Valiant s BulkSynchronous Parallel[1] model. Parallel languages are often written as a ....

....for Abstract Data Types. We can achieve this using object oriented programming. As object oriented parallel languages already exist, I must first justify the need for a new one. The most important reason is that those which do exist are complicated and not BSP oriented. DoPVM, Mentat[7] and PC [8] are three commonly cited languages: These are essentially different (as is my language) DoPVM is a fundamentally different approach to parallelism and is not suited to BSP. Mentat and PC are both SPMD with automatic compiler included synchronisation. Mentat and PC are similar to existing ....

[Article contains additional citation context not shown here]

J. Lee & D. Gannon, Indiana University,Object-Oriented Parallel Programming Experiments and Results.

....results. 4.1 Parallel Object Oriented Systems In the object oriented parallel processing domain Mentat differs from systems such as [13] 14] shared memory C systems) in its ability to easily support both shared memory MIMD and distributed memory MIMD architectures, as well as hybrids. PC [15] and Paragon [16] on the other hand are dataparallel derivatives of C . Mentat accommodates both functional and data parallelism, often within the same program. ESP [17] is perhaps the most similar of the parallel object oriented systems. It too is a highperformance extension to C that supports ....

J. K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, pp. 273-282, Albuquerque, NM, 1991.

....memory, MIMD 1. Introduction The object oriented paradigm has proven to be a powerful tool for managing complexity in the development of software for sequential computers, and its power is being exploited in the more complex domain of software for parallel machines as well [5] 6] 8] 22][32][40] Software for parallel machines must provide efficient run time support if it is to achieve acceptable performance. This is particularly true for distributed memory MIMD machines where communication and synchronization costs can be large. Given that high performance is the raison d etre of ....

....Mentat objects and their contained data elements. To eliminate the need to hand generate this rather tedious code we are exploring the incorporation of data parallel language features into the Mentat programming language [46] The language extensions are similar to those being developed by Gannon [32], Quinn [28] Larus [31] and others. Our work is significantly different in that the result will be a language that combines task and data parallelism in a seamless fashion, while the above languages are data parallel only. 3. The Run Time System (RTS) MPL programs are executed on a virtual ....

J. K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, pp. 273-282, Albuquerque, NM, 1991.

....HPF programs are architecture independent although it is likely that the data mapping directives may need to be architecture specific to achieve reasonable performance. HPC , the emerging C equivalent to HPF, is based on a unification of two C based parallel languages CC [24] and pC [17]. CC , developed at CalTech, is C extended with single assignment variables, parallel composition and atomic functions. Single assignment variables enable synchronization between distinct activities. Parallel composition enables multiple C statements (which could be function calls) to be ....

J-K Lee and D. Hannon. Object oriented parallel programming: Experiments and results. In Proceedings of SUPERCOMPUTING'91, November 1991.

....gaining more and more popularity for their optimal scalability. However, programming them by using a message passing library is still difficult, architecture dependent and error prone. Therefore, data parallel programming languages such as HPF [7] High Performance Fortran) Fortran D [3] PC [9], MPC [8] PROMOTER [5] and many others have been introduced to support machine independent expression of data parallelism. Efficient implementation of data parallel languages is especially difficult, considering that these languages attempt to provide a missing link between data parallel ....

Lee, J. K. and Gannon, D.: Object-oriented Parallel Programming: Experiment and results. Proc. of Supercomputing 91, 273-282, Nov. 1991.

....If successful, Fortran D and High Performance Fortran may make it possible to maintain a single image of a Fortran program for a variety of parallel machines. A research project led by Dennis Gannon at Indiana University is developing pC , which addresses the same problem for C programs [36]. When combined with process based methods for specifying task parallelism, these languages and their associated compiler technology may represent a first step toward a truly comprehensive solution to the portable parallel programming support. Nevertheless, there is still much to do. Compiler ....

J. K. Lee and D. Gannon. Object oriented parallel programming: experiments and results. Proceedings of Supercomputing 91, pages 273--282, Albuquerque, New Mexico, November 1991.

....at once, or 4) Allow the sender of a message to proceed in parallel with the receiver. The new generation of parallel object oriented languages supports inter object parallelism as well as intra object parallelism. We think that data parallel object oriented languages such as CC , C or PC [12] achieve the goal of high expressive power and ease of programming by allowing programmers to use data parallel constructs inside objects. Their importance will certainly increase in the future, as the community of parallel systems users grows from exclusively Fortran educated scientists and ....

J. K. Lee, Dennis Gannon, "Object-Oriented Parallel Programming Experiments and Results", Proceedings of Supercomputing '91, Alberquerque, NM, November 18-22, IEEE Computer Society Press, Los Alamitos, California, 1991.

....languages. In Mentat[11] developed by Grimshaw and his research group, the user can specify the granularity and partitioning decisions using extended C while the compiler and run time system manage communication, synchronization, and scheduling. There are many other systems and projects[18,19,20] that are similar in some respects to Mentat, since an easy way to provide concurrent object oriented programming systems might be to augment C with communication and synchronization libraries that must be explicitly used by programmers. However, this approach does not offer the computational ....

J.K. Lee and D. Gannon, "Object-Oriented Parallel Programming Experiments and Results, " Proc. Supercomputing 91, IEEE CS Press, Los Alamitos, Calif., Order No. 2159-02, pp.273-282.

....parallel processing 1. Introduction The object oriented paradigm has proven to be a powerful tool for managing complexity in the development of software for sequential computers, and its power is being exploited in the more complex domain of parallel software as well [3] 4] 6] 14][22][23] In parallel systems, where performance is a key objective, the issue of run time support must be addressed, particularly for distributed memory MIMD machines. For years the object oriented approach has had a reputation for high run time overhead and poor performance. Recent sequential ....

J. K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, pp. 273-282, Albuquerque, NM, 1991.

....by the seemingly unpredictable results caused by the intricacies of data dependences in their programs. A group at Indiana University developed an object based extension of Concurrent Aggregates, called Distributed Collection Model, and implemented the model in a language called pC based on C [GL92] pC programs run very effectively on high performance parallel architectures such as the CM 5. pC programs could be compiled to execute effectively on workstation networks, but the language model currently is not capable of implementing recursive data structures or nested parallelism. ....

Dennis Gannon and Jeng-Kuen Lee. Object-oriented parallel programming: experiments and results. In Proceedings of Supercomputing '91. IEEE, 1992.

....language pC . The focus has been on a complete discussion of the language extensions and the associated semantics as well as providing an introduction to the standard library of linear algebra collection classes. We have not cited any performance results here. These are described in other papers [7], 9] 11] 13] and new results for various benchmarks will be available soon. The compiler currently runs on any standard scalar unix system as well as the Thinking Machines CM 5. A version for the Intel Paragon will be available by the middle of 1993. A group led by Jenq Kuen Lee at NTHU in ....

J. K. Lee and D. Gannon, Object Oriented Parallel Programming: Experiments and Results Proceedings of Supercomputing 91 (Albuquerque, Nov.), IEEE Computer Society and ACM SIGARCH, 1991, pp. 273--282.

....references in the conjugate gradient and bitonic sort programs written in pC . Conjugate gradient is one of the programs in the pC test suite. It solves the equation A u = f for u, given f by a conjugate gradient method. The details of the pC implementation are described elsewhere[10]. The main distributed data structure is the Grid collection that represents the sites in a two dimensional grid. The individual elements at each site of the Grid collection are small 2D array s. Figure 5 shows the distributed data structure references. We record the distributed data structure ....

J. K. Lee and D. Gannon. Object oriented parallel programming: Experiments and results. In Proceedings of Supercomputing 91 (Albuquerque, Nov.), pages 273--282. IEEE Computer Society and ACM, 1991.

....of the target by the run time system. ffl A function object called the Alignment. This function maps collection elements to the abstract coordinate system of the Distribution object. The pC language has a library of standard collection classes that may be used (or subclassed) by the programmer [36, 49, 17, 20]. This includes collection classes such as DistributedArray, DistributedMatrix, DistributedVector, and DistributedGrid. To illustrate the points above, consider the problem of creating a distributed 5 by 5 matrix of floating point numbers. We begin by building a Distribution. A distribution is ....

J. K. Lee and D. Gannon. Object oriented parallel programming: Experiments and results. In Supercomputing 91 (Albuquerque, Nov.), pages 273--282. IEEE Computer Society and ACM, 1991.

No context found.

Dennis Gannon and Jeng-Kuen Lee. Object-Oriented Parallel Programming: Experiments and Results. In Proceedings of Supercomputing '91 . IEEE, 1991.

No context found.

J. K. Lee and D. Gannon, "Object-oriented parallel programming: experiments and results," in Proceedings of Supercomputing '91, 1991, pp. 273-282.

No context found.

J. K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, pp. 273-282, Albuquerque, NM, 1991.

No context found.

J. K. Lee and D. Gannon, "Object-Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing '91, IEEE Computer Society Press, 1991.

No context found.

J. K. Lee and D. Gannon, "Object Oriented Parallel Programming Experiments and Results," Proceedings of Supercomputing `91, pp. 273-282, Albuquerque, NM, 1991.

No context found.

J.K. Lee and D. Gannon. Object Oriented Parallel Programming: Experiments and Results. In Proceedings of Supercomputing '91. IEEE Computer Society and ACM SIGARCH, 1991.

First 50 documents

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