Ueda, K. and Chikayama, T. 1990.Design of the kernel language for the parallel inference machine. Computer Journal 33, 6, pp.494-500.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and has the notion of guarded clause and committed choice. In this language, the system does not search solutions by backtracking. Therefore, the value of a variable once computed is never revoked. In the following sections, we will study implementations of IM2 machines in the language KL1[11], which is based on the concept of Flat GHC [12,13] Since Flat GHC is a sublanguage of GHC, we will simply call the language GHC. A GHC program is a set of guarded clauses of the following form: H : G 1 ; G n B 1 ; Bm 12 main : pinf(Y) sum(X,Y) naturals(0,X) ....

Kazunori Ueda and Takashi Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, 33(6):494-500, 1990.

....of programs as the simplest program specification in addition to mode and type analyses, which are the simplest technique of program analysis. Although the framework is quite general, whether it is practical or not may depend on the choice of a language. We targets the error correction of KL1 [34] programs assuming strong moding and typing of Moded Flat GHC. KL1 is designed based on Flat GHC that is not equipped with strong moding typing, but the debugging of KL1 programs turns out to benefit from moding and typing. Furthermore, its compiler KLIC provides a nice platform for our ....

K. Ueda and T. Chikayama. Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, 33(6):494--500, 1990.

....variables) as well as the normal variablelocking mechanism. Only one variable is locked at a time to avoid complications due to locking an entire set of variables at once (compare to Section 7) The system GDCC was implemented on top of KL1, a committed choice language developed at ICOT (see [35, 36]) The system allows the user to plug in constraint solvers via a so called stream interface, which handles the synchronization between the parallel running constraint solvers and the inference machine. There are three constraint solvers available: algebraic, boolean, and linear, all using ....

K. Ueda and T. Chikayama. Design of the kernel language for the parallel inference MAchine. Computer Journal, 33(6):494--500, December 1990.

....case of Prolog, a transformational language, a program can be read declaratively as a description of the final result that it computes: provided that the program terminates, its result is guaranteed to satisfy this specification. Concurrent logic programming languages such as Parlog [10] and KL1 [16] are declarative in the same sense as Prolog: a program can explicitly specify its final result. Unfortunately, all other (both safety and progress) properties of a program remain implicit: they have to be preserved by proper use of control features (modes, sequencing, etc. Many concurrent logic ....

Ueda, K. and Chikayama, T. Design of the kernel language for the Parallel Inference Machine. Computer Journal 33, 6, pp. 494-500, 1990. 15

....Since different type of slices may provide different information for users, in this paper, we will consider to compute executable slices for concurrent logic programs. 3 Preliminaries We assume that readers are familiar with the basic concepts of logic programs, and in this paper, we will use KL1 [9], a concurrent logic programming language based on Guarded Horn Clauses (GHC) as our target language. This language illustrates the basic mechanisms of concurrent logic programming. A term is a variable, a constant, or a compound term f(t 1 ; t n ) where f is an n ary function symbol and ....

K. Ueda and T. Chikayama, "Design of the Kernel Language for the Parallel Inference Machine," The Computer Journal, Vol.33, No.6, pp.494-500, 1990.

....of our program analysis system that supports the development of software engineering tools for concurrent logic programs, and conclusions are given in Section 6. 2 Preliminaries We assume that readers are familiar with the basic concepts of logic programs, and in this paper, we will use KL1 [18], a concurrent logic programming language based on Guarded Horn Clauses (GHC) as our target language. This language illustrates the basic mechanisms of concurrent logic programming. A term is a variable, a constant, or a compound term f(t 1 ; t n ) where f is an n ary function symbol and ....

Ueda, K. and Chikayama, T. :Design of the Kernel Language for the Parallel Inference Machine, The Computer Journal, Vol.33, No.6, pp.494-500 (1990).

.... BNs JB9T=hM r8 8l W j F # V G 5 ] H 9 k W m 0 i s 08 8l G O 3 l i B9T 3 s F 9 H N )8f5 9= rMQ0U 7 F k #Nc ( P JBNs V 8 H;X8 8 8l ABCL[14] 18] G O future 9=J8 G W m ; 9 9 l C I N 8 . touch9=J8 GF14 r 5 ] H 7 F k #GHC[16] d GHC r85 K 7W 5 l JBNsO M 7 8 8l KL1[3] G O J# t N a a N G # N 4 k72 ,JB9T JBNs KJL W m ; 9 H 7 FF0:n 7F k # W m ; 94V NF14 O pJs N6 M jCJ G b k 9 H j rDL 8 DL . K h C F B8= 5 l k #B K O 9 l C I N 8 . F14 N a K fork join 7 N9=J8 rMQ0U 7 F k8 8l b k # 2.4.2 Java 8 8l Java ....

Kazunori Ueda an T. Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, Vol. 33(6), pp. 494--500, 1990.

....model and the mathematically tractable semantics of nonprocedural languages. Unnatural extensions can lose the raison d etre of nonprocedural languages and may well hamper sophisticated optimization. KL1, designed based on Flat GHC, already provided array features based on this principle [7]. At the language level, KL1 provided array operations within the framework of logic programming, while at the implementation level, one bit reference counting was employed for the efficient update of single reference arrays [2] They achieved the same time and space complexities as those of ....

.... of regular problems because: 8 ffl the mode system provides both programmers and implementation with a useful fundamental information by analyzing the handling of data as resources , and ffl the conceptual clarification they provide (e.g. the separation of concurrency and parallelism [7]) should be useful in any future applications. In particular, the possibility of being able to control aliasing much more nicely than in procedural parallel programming is worth exploring both for better programming styles and for program optimization. ....

Ueda, K. and Chikayama, T., Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (December 1990), pp. 494--500.

....of KL1 and program dependences in concurrent logic programs. Section 3 gives an overview of system CLPKIDS. Concluding remarks are given in Section 4. 2 Background 2. 1 KL1 We assume that readers are familiar with the basic concepts of logic programs, and in this paper, we will use KL1 [14], a concurrent logic programming language based on Guarded Horn Clauses (GHC) as our target language. This language illustrates the basic mechanisms of concurrent logic programming. A term is a variable, a constant, or a compound term f(t 1 ; t n ) where f is an n ary function symbol and ....

K. Ueda and T. Chikayama, "Design of the Kernel Language for the Parallel Inference Machine," The Computer Journal, Vol.33, No.6, pp.494-500, 1990.

....the application specific code is abstracted away, and hence can be written in any programming language. 1.1 Related work Concurrent logic programming and concurrent constraint programming is an important influence for this work. Both concurrent logic programming languages (e.g. Parlog [4] KL1 [16]) and concurrent constraint programming languages [15] use a central constraint store for agent synchronization and communication and preserve many of the benefits of the abstract declarative logic programming model, such as the logical reading of programs and the use of logical terms to represent ....

Ueda, K. and Chikayama, T. 1990. Design of the kernel language for the parallel inference machine. Computer Journal 33, 6, pp.494--500. 11

....of a functional language Haskell, the program comes to have di erent semantics under the usual lazy evaluation strategy. We have implemented this Gray code input output mechanism using logic programming languages. We have written gtos, stog, and the addition function pl of Section 8 using KL1 [UC90], a concurrent logic programming language based on Guarded Horn Clauses. We have also implemented them using the coroutine facility of SICStus Prolog. We are also interested in extending lazy functional languages so that programs in Section 7 and 8 become executable. The details about these ....

Kazunori Ueda and Takashi Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, 33(6):494-500, 1990.

.... 0 = DNil D2 (olist 0) olist 0) def dappend DNil DNil = DNil dappend (D2 A1 A2) DNil = D2 A1 A2 dappend DNil (D2 B1 B2) D2 B1 B2 dappend (D2 A1 A2) D2 B1 B2) A2.OCons2 = B1 In D2 A1 B2 ; Readers familiar with parallel logic programming languages (such as KL1, [43]) will perhaps by now have seen a similarity between I structures and logic variables and their use as a way to communicate data from a producer to multiple consumers. In fact, the connection is deep, and has been explored in detail by Jagadeesan and Pingali [21] If I structure assignment is ....

....which restrict the OR choice to a small, deterministic conditional test. Further, some languages use mode declarations (programmer or compiler supplied) to specify the direction of data flow during unification, so that unification can be simplified substantially. An example is the language KL1 [43] that is used in the Japanese Fifth Generation project. We believe that such restricted logic languages are, in fact, not very different from functional languages. Further, the simple, one way data flow use of logic variables is very similar to the use of I structures in Id. Thus, at an abstract ....

K. Ueda and T. Chikayama. Design of the Kernel Language for the Parallel Inference Machine. The Computer J., 33(6):494--500, 1990.

....optimal query tree. Also, the active database can easily obtain information for optimization by sampling databases. 2 Parade: A Parallel Active Database Prototype 2. 1 Overview of the System Parade is a parallel active relational database prototype programmed with a parallel logic language, KL1 [17]. Due to KLIC [7] a translator from KL1 to C, Parade is easily portable to many parallel computers like an nCUBE2 and work station clusters connected with LAN. The KL1 provides an easy way to describe fine grain, concurrent, and message oriented processes. The model of the KL1 processes are ....

K. Ueda and T. Chikayama. Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, 33(6):494--500, Jun 1990.

....threads significantly enhance the maintainability of parallel programs, which is increasingly becoming an important issue as parallel platforms are widespread. Fine grain multithreaded systems have typically been provided in programming languages that have a builtin notion of parallelism [1, 4, 12, 18, 26, 29, 30], whereas coarse grain systems, such as Pthreads and Solaris threads, are available as libraries accessible from standard sequential languages, whose compilers are not necessarily aware of multiple threads of control. This is not an accident, because the underling execution mechanism of fine grain ....

Kazunori Ueda and Takashi Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, 33(6):494--500, 1990.

....threads significantly enhance the maintainability of parallel programs, which is increasingly becoming an important issue as parallel platforms are widespread. Fine grain multithreaded systems have typically been provided in programming languages that have a builtin notion of parallelism [1, 4, 10, 15, 23, 25, 26], whereas coarse grain systems, such as Pthreads and Solaris threads, are available as libraries accessible from standard sequential languages, whose compilers are not necessarily aware of multiple threads of control. This is not an accident, because the underling execution mechanism of fine grain ....

Kazunori Ueda and Takashi Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, 33(6):494--500, 1990.

.... was born in early 1980 s from the process interpretation of logic programs [34] Relational Language [7] the first concrete proposal of a concurrent logic language, was followed by a succession of proposals, namely Concurrent Prolog [20] PARLOG [8] and Guarded Horn Clauses (GHC) 27] KL1 [29], the Kernel Language of the Fifth Generation Computer Systems (FGCS) project [22] was designed based on GHC by featuring (among others) mapping constructs for concurrent processes. To be precise, KL1 is based on Flat GHC [28] a subset of GHC that restricts guard goals to calls to test ....

.... parallel programming languages, it achieves clear separation of concurrency (concerned with logical aspects of programs) and parallelism (concerned with physical mapping of processes) We regard this separation of concerns as the most important achievement of KL1 and its parallel implementation [29]. In other words, by using logical variables as communication channels we had achieved 100 network transparency within system area networks (SAN) The fact that programs developed and tested on sequential machines ran at least correctly on parallel machines has benefited us enormously in the ....

Ueda, K. and Chikayama, T. Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....the sense of the # calculus. When the author proposed Guarded Horn Clauses (GHC) 36] 37] as a simplification of Concurrent Prolog and PARLOG [8] the principal design constraint was to retain channel mobility and evolving process structures [32] because GHC was supposed to be the basis of KL1 [39], a language in which to describe operating systems of the Parallel Inference Machines as well as various knowledge based systems. 4. Non strict. Logical variables provide us with the paradigm of computing with partial information. Interesting programming idioms including short circuits, ....

Ueda, K. and Chikayama, T. Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....and the design of KL1 started based on Flat GHC. While GHC was designed as a concurrent language that did not address how programs should be executed, KL1 was designed as a parallel language in which programmers could specify what processor should execute what processes and at what priorities [13]. For the detailed history of the kernel language design in the Fifth Generation Computer Project, the readers are referred to [14] 1 Several more concurrent logic programming languages were proposed in the past decade, but the difference between all those languages is rather small as 1 An ....

Ueda, K. and Chikayama, T., Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....is language defined. Flat GHC is different from GHC in that guard goals are restricted to ffl unification goals and ffl calls to test predicates, namely predicates defined by clauses with empty bodies. However, this restriction is not really restrictive as a process description language [20]. We have found that it is more natural to distinguish between test predicates that define rewrite conditions and non test predicates that define processes; they have very different characteristics. Thus, in this paper, we make a clearer distinction by assuming that predicates (other than the ....

Ueda, K. and Chikayama, T., Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....unification. A unification body goal of the form t 1 = t 2 can cause bidirectional dataflow in general, but mode analysis tries to guarantee that at least one of t 1 and t 2 is an uninstantiated variable and hence the goal does not fail except due to occur check. Our experience with GHC and KL1 [Ueda and Chikayama 1990] has shown that the full functionality of bidirectional unification is seldom used and that programs using it can be rewritten rather easily, if not automatically, to programs using unification as assignment. 2 These languages are indeed used as general purpose concurrent languages, which ....

....0.81 1.71 (294 kRPS) 4 PEs 0.44 0.63 1.33 (377 kRPS) 5 PEs 0.36 0.53 1.10 (456 kRPS) 6 PEs 0.33 0.46 0.96 (523 kRPS) 7 PEs 0.33 0.39 0.85 (591 kRPS) 8 PEs 0.33 0.36 0.77 (652 kRPS) C (recursion) cc O 0.71 0.72 C (iteration) cc O 0.32 0. 35 ( 3 kilo Reductions Per Second) construct of KL1 [Ueda and Chikayama 1990], but it should scarcely affect the result of performance evaluation described below. Locking of shared resources, namely logic variables, goal records, communication cells, the global deque, etc. is done using the xchg (exchange) instruction as usual. Using Program 1, we measured the ....

K. Ueda and T. Chikayama, Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (Dec., 1990), pp. 494--500. -- 16 --

....from more than one place, some of which may be recursive calls. The same mode constraint may be imposed by different calls. Although the framework is quite general, whether it is practical or not may depend on the choice of a language. Kima corrects wrong occurrences of variable symbols in a KL1 [7] program assuming strong moding and typing of Moded Flat GHC. KL1 is designed based on Flat GHC that is not equipped with strong moding typing, but the debugging of KL1 programs turns out to benefit from moding and typing. Furthermore, its compiler KLIC provides a nice platform for our experiments ....

Ueda, K. and Chikayama, T., Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....from more than one place, some of which may be recursive calls. The same mode constraint may be imposed by di#erent calls. Although the framework is quite general, whether it is practical or not may depend on the choice of a language. Kima corrects wrong occurrences of variable symbols in a KL1 [7] program assuming strong moding and typing of Moded Flat GHC. KL1 is designed based on Flat GHC that is not equipped with strong moding typing, but the debugging of KL1 programs turns out to benefit from moding and typing. Furthermore, its compiler KLIC provides a nice platform for our experiments ....

Ueda, K. and Chikayama, T., Design of the Kernel Language for the Parallel Inference Machine. The Computer Journal, Vol. 33, No. 6 (1990), pp. 494--500.

....processing power with a comfortable software development environment based on concurrent logic programming. The parallel inference system consists of parallel inference machines [7] parallel implementations of a concurrent logic programming language KL1 upon them and basic software systems [17, 2]. Through development of numerous experimental parallel application software systems, the parallel inference system was proved to provide superior processing power and a comfortable environment for efficient software development. However, the system based on special purpose hardware had some ....

Kazunori Ueda and Takashi Chikayama. Design of the kernel language for the parallel inference machine. The Computer Journal, 33(6):494--500, December 1990.

No context found.

Ueda, K. and Chikayama, T. 1990.Design of the kernel language for the parallel inference machine. Computer Journal 33, 6, pp.494-500.

No context found.

Ueda, K. and Chikayama, T. 1990. Design of the kernel language for the parallel inference machine. Computer Journal 33, 6, pp.494-500.

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