Gary W. Sabot. The Paralation Model Architecture-Independent Parallel Programming. The MIT Press, Cambridge, Massachusetts, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....AI program, data parallelism is very easy to identify. Such parallelism can be exploited in a language where primitive functions operate directly on the aggregate data, rather than on one datum at a time. Such primitive functions can be mapped directly onto vector or certain parallel hardware [5]. Scheme has several advantages over C for data parallel programming. Algorithms expressed as C programs cannot easily make use of such parallelism, since C programs are in a sense already compiled by the programmer for the serial machine described by C (i.e. a PDP 11) Scheme is extensible: ....

....is determined at run time. While Elk is interpreted, we wanted to write code which could also be compiled using an appropriate redefinition of parlet. The parlet form declares all vector variables and all unknown vector functions. The parlet form resembles the elwise form used in Paralation Lisp [5]; both forms are unusual for Lisp like languages in that they make a type distinction explicit. The syntax of parlet is (parlet declarations body ) As a simple example, parlet (v) v 1) declares v to be a vector and returns the result of comparing 1 to each element of this vector. ....

Gary Sabot. The Paralation Model: ArchitectureIndependent Parallel Programming. MIT Press, Cambridge, Mass., 1988.

....process, per window. 2.1.3 High Level Parallel Programming Models and Languages SIMD languages such as HP Fortran [30] Lisp [42] and APL [20] provide data parallel primitives which are a thin veneer on standard vector parallel operations. Idealized SIMD languages such as Paralation Lisp [60] and NESL[8] express parallelism with parallel apply operations performed over data collections (typically vectors) nested parallelism is allowed, but is compiled into non nested, vector parallel operations. None of these languages is well suited to distributing objects non uniformly over a ....

Gary W. Sabot. The paralation model: architecture-independent parallel programming. MIT Press, 1988.

....means that there is an O(N) storage requirement per process, per window. guages SIMD languages such as HP Fortran [30] Lisp [42] and APL [20] provide data parallel primitives which are a thin veneer on standard vector parallel operations. Idealized SIMD languages such as Paralation Lisp [60] and NESL[8] express parallelism with parallel apply operations performed over data collections (typically vectors) nested parallelism is allowed, but is compiled into non nested, vector parallel operations. None of these languages is well suited to distributing objects non uniformly over a ....

Gary W. Sabot. The paralation model: architecture-independentparallelprogramming. MIT Press, 1988.

....vector and matrix computations. However, its prefix and communication statements hide surprising power and versatility, enabling it to express algorithms in such diverse areas as graph theory, computational geometry, pattern matching, logic inference and combinatorial optimization. In fact, Sabot [Sab88] mentions that in a survey of 120 parallel algorithms from three ACM Symposia on Theory of Computing (STOC) all of them were found to naturally fit the data parallel 18 model. Perhaps the reason for this popularity is that a data parallel program has a single, sequential thread of control; ....

Gary Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, Cambridge, MA, 1988.

....the physical hardware on which a program might run, since the connection between a language level model of parallelism and its supposed obvious implementation platform is rather tenuous. For example, Hatcher and Quinn [55] describe a data parallel language compiler for MIMD machines, while Sabot [127] describes how an SIMD computer could be used to simulate an MIMD computer. 2.1.4 Control parallelism Control parallelism is a form of functional parallelism characterised mainly by explicit communication ( message passing ) and synchronisation between processes. In effect, the programmer writes ....

Gary Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, 1988.

....the physical hardware on which a program might run, since the connection between a language level model of parallelism and its supposed obvious implementation platform is rather tenuous. For example, Hatcher and Quinn [55] describe a data parallel language compiler for MIMD machines, while Sabot [127] describes how an SIMD computer could be used to simulate an MIMD computer. 2.1.4 Control parallelism Control parallelism is a form of functional parallelism characterised mainly by explicit communication ( message passing ) and synchronisation between processes. In effect, the programmer writes ....

Gary Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, 1988.

....C (Rose Steele Jr. 1987) Dataparallel C (Hatcher Quinn, 1991) and High Performance Fortran (Forum, 1993) There are also a number of proposals for functional data parallel languages such as Connection Machine Lisp (Wholey Steele Jr. 1987) Lisp (Lasser, 1986) Paralation Lisp (Sabot, 1988), TUPLE (Yuasa, 1992) Plural EuLisp (Merrall Padget, 1992) DPML (Hains Foisy, 1993) Nesl (Blelloch, 1993) FX (Talpin Jouvelot, 1993) and Caml Flight (Foisy Chailloux, 1995) These languages enable us to exploit data parallelism in functional programming, but the source of ....

Sabot, G. (1988). The paralation model: Architecture-independent parallel programming. Cambridge, MA: MIT Press.

....environment and a parallel programming model which is general enough to handle the diversity inherent in such an environment. The model must be architecture independent, must be intuitive, simple to use and must be accurate enough to reflect its cost of execution. The Paralation Model presented in [14] attempts to meet these requirements. 2.2 Portable Application Description Medium A number of application description media, capable of describing and handling parallelism have been proposed. However, they either lack the flexibility to exploit the potential of the underlying hardware (e.g. ....

Gary W. Sabot, The Paralation Model: Architecture-Independent Parallel Programming, The MIT Press, Cambridge, Massachusetts, 1988.

....in parallel with the others. Clearly the traditional execution model underlying flat data parallelism cannot provide such parallelism within parallelism . Thus, to directly support nested data parallelism we need to consider generalisations of that execution model. Several language systems [23, 17] which allow nested data parallel specifications choose to solve this problem by arbitrarily serialising all but one dimension of parallelism. This effectively re establishes the property that only a single data parallel operation may be active within the system at any time. Thus such ....

G. Sabot. The Paralation Model: Architecture Independent Parallel Programming. The MIT press, Cambridge Massachusetts, 1990.

....is merged into the global state. Each private variable has its values in all processes combined using a specified merge function f , and the result updates the corresponding variable in the enclosing scope. This combining action is similar to that used to resolve conflict in message collisions [Sab88] although Proteus applies the reduction of f only across the changed values from all processes. While in our examples the merge function defaults uniformly to arbitrary selection, Proteus also permits merge functions to be individually specified for each variable through declarations of the ....

....might be roughly divided into the following categories. ffl Languages with widely translatable logical models, such as Linda s distributed data structures [CGL86] the synchronization variable methods of Strand [FT90] and PCN [CT92] or the data parallel abstraction of the Paralation model [Sab88] Often called coordination languages, these form a harness in which programs in different sequential computation languages can cooperate in parallel [CG91] ffl Languages which incorporate a large variety of parallel primitives, such as Ease [Zen90] ffl Wide spectrum parallel languages that ....

G. Sabot, The Paralation Model: ArchitectureIndependent Parallel Programming. MIT, 1988.

....that every merge operation is to apply, for each private variable, the reduction of the binary operation f across the ordered sequence of all processes values, yielding the global update. This is similar in spirit to other uses of combining functions to resolve conflict in message collisions [Sab88] When combining a variable s value from each of the processes, a key consideration is whether the variable has changed since the last merge operation. Consequently we define the combining function to apply only between the changed values. Formally, for the combining operation we augment the ....

....which might be roughly divided into the following classes. ffl Languages with widely translatable logical models, such as Linda s distributed data structures [CGL86] the synchronization variable methods of Strand [FT90] and PCN [CT90] or the data parallel abstraction of the Paralation model [Sab88] ffl Languages which incorporate a large variety of parallel primitives, such as Ease [Zen90] and Alloy [MH90] ffl Wide spectrum parallel languages that rely on refinement from architecture independent specification. Notable wide spectrum parallel language efforts include Crystal [Che86] and ....

G. Sabot, The Paralation Model: ArchitectureIndependent Parallel Programming. MIT Press, 1988.

....particularly important for parallel programming, in part because control more directly affects performance. Yet, to our knowledge, only those parallel programming languages that inherit control abstraction from a parent sequential language support it. For example, both Multilisp [6] and Paralation [12] use Lisp closures in the implementation of the parallel programming constructs presented to users, but their developers have not argued the benefits of control abstraction as a parallel programming tool. One primary benefit of control abstraction is that it separates the use of control from its ....

G. W. Sabot. The Paralation Model: ArchitectureIndependent Parallel Programming. MIT Press, 1988.

....performing computations on a large set of data. This is perhaps the reason for the preponderance of data parallelism in massively parallel computations. For example, all of the more than 120 parallel algorithms in a survey of three ACM Symposia on Theory of Computing (STOC) were data parallel [17, 18]. One way of executing a data parallel program is on an SIMD (Single Instruction Multiple Data) computer. The semantics of the abstract model have often been confused with its hardware implementation, and data parallelism has been equated to SIMD. However, not only is it possible to execute a ....

Gary Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, Cambridge, MA, 1988.

....computation communities, even for programming serial machines. There are data parallel versions of the most popular serial languages C (C [110, 133] Dataparallel C [60] Fortran (Fortran 90 [13] CM Fortran [134, 115] Fortran D [55] and others) and Lisp (CM Lisp [126] Paralation Lisp [113, 114], Lisp [92, 135, 136] as well as data parallel languages designed from first principles (NESL [17] VCODE [18] and sequential languages that were already partly data parallel (APL [76] and its dialects [73, 77, 80, 142] SETL [41, 116] etc. This list is far from exhaustive. Data ....

Gary W. Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, Cambridge, Massachusetts, 1988.

....programming, cooperative processing and task parallelism. Its focus is synchronization and communication between processes. While it is object oriented, it would be difficult to build data parallel collections with it. There is no support for nested parallelism. Paralation Lisp Paralation Lisp [16] supports collections in what are called paralations (parallel relations) A paralation has a shape ( vector or array, for example) and a number of data elements called fields. Paralation Lisp adds parallel collections to Common Lisp along with an elwise clause that is nearly identical in ....

Gary W. Sabot. The Paralation Model: Architecture-Independent Parallel Programming. Series in Artificial Intelligence. MIT Press, 1988.

....with other compiled modules. Many other projects are building parallel programming systems that support nested parallelism, object oriented programming, or collection oriented programming. Examples include pC [6] CC [7] Concurrent Aggregates [9] C [10] Proteus [11] and Paralation Lisp [14]. For a detailed comparison of these projects see the technical report version of this paper available from the authors. However, no other project currently addresses the integration of nested data parallelism into an object oriented language. 1.1 Organization of this paper The rest of this ....

....polymorphism and the more ad hoc operator overloading are used in building these classes. Stream I O is cleanly integrated into the language. Finally, static type checking is used extensively to assist in compile time optimizations. This final feature differentiates our effort from Paralation Lisp [14], which relies on dynamic type checking. Two major features of C that we have not used so far are inheritance and exception handling. We feel that these features appear when writing large codes, independent of whether the code is sequential or parallel. Thus, we expect these constructs to show ....

G. W. Sabot. The Paralation Model: ArchitectureIndependent Parallel Programming. Series in Artificial Intelligence. MIT Press, 1988.

....chosen a data parallel style of programming, where parallelism is expressed as (parallel) operations over (large) sets of data. A large fraction of the existing parallel algorithms for PRAM and other machine models [46] are either data parallel in nature or can be easily converted to such a form [9, 32, 58]. Data parallel languages have historically been linked with SIMD parallel computers, and researchers have largely shied from implementing such languages on MIMDparallel machines. A nave implementation of data parallelism on a MIMD machine has the following performance bottlenecks, which affect ....

SABOT,G.W. The Paralation Model: Architecture-Independent Parallel Programming. The MIT Press, Cambridge, MA, 1988.

.... using less trivial programmer defined associative functions in computing reductions of vectors is also provided in several languages for parallel computing, such as iPSC 2 Fortran and C [Int89] and the innovative and less conventional languages Connection Machine Lisp [SH86] and Paralation Lisp [Sab88]. Within conventional semantic models, the level of abstraction provided by such general reduction operators is less than ideal when nontrivial programmer defined functions are used. This is due to such deficiencies as the lack of a simple, implementation independent conceptualization for the ....

....fact that diva procedures can have several dynamic parameters permits space utilization to be appropriate to a particular usage, within a conventional semantic model, including the use of a vector of records when appropriate. This storage advantage is also provided by the fields of a paralation [Sab88]: such fields can be added and deleted according to the needs of different parts of the program. The semantics of the noteworthy paralation model are much richer and more powerful than conventional semantics. 10 Similarly, the implementation of diva calls involving Z on a distributed memory ....

[Article contains additional citation context not shown here]

Sabot, G. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, Cambridge, 1988.

....the reduction R(f; s) of a finite sequence s using the function f . Such a general reduction operator is provided in several languages for parallel computing, such as iPSC 2 Fortran and C [Int89] and the innovative and less conventional languages Connection Machine Lisp [HS86] and Paralation Lisp [Sab88]. Since R need apply f only to the sequence s, it is not necessary that f be associative, only that f satisfy an associativity property relative to the sequence s. The concept of associativity relative to a sequence is defined and studied in this paper. This concept extends the applicability of R; ....

Sabot, G. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, Cambridge, 1988. 21

....they can contain aggregates as values. The expressive utility of nested data parallelism was recognized long ago in high level sequential programming languages like SETL [16] and APL2 [11] Parallel execution of nested data parallelism has been realized by recent languages such as Paralation Lisp [15], NESL [3] and Proteus [7,9] In Proteus, our high level, wide spectrum parallel language, all data parallelism is expressed using an iterator construct which is analogous to the comprehension construct of set theory. For example, if A is a sorted sequence of integers and S is an arbitrary ....

G . Sabot, The Paralation Model : ArchitectureIndependent Parallel Programming, MIT Press, 1988.

....of the Connection Machine 2. These often included a map operation, some form of reduction, perhaps using only a fixed set of operators, and later scans (parallel prefixes) and permutation operations. In approximately chronological order, these models are: scan [32] multiprefix [170] paralations [100, 171], the C data parallel language [111, 165] the scan vector model and NESL [33 38] and CamlFlight [109] As for other data parallel languages, these models are simple and fairly abstract. For instance, C is an extension of the C language that incorporate features of the SIMD parallel model. In ....

G. Sabot. The Paralation Model: Architecture-Independent Parallel Programming. MIT Press, 1989.

No context found.

Gary W. Sabot. The Paralation Model Architecture-Independent Parallel Programming. The MIT Press, Cambridge, Massachusetts, 1988.

No context found.

Gary W. Sabot. The Paralation Model: Architecture-Independent Parallel Programming. The MIT Press, 1988.

No context found.

Gary W. Sabot. The Paralation Model: Architecture-Independent Parallel Programming. The MIT Press, Cambridge, MA, 1988.

No context found.

Sabot, G. W. The Paralation Model: Architecture-Independent Parallel Programming.The MIT Press, Cambridge, Massachusetts, 1988.

First 50 documents  Next 50

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