K. Ueda and M. Morita. A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming (ICLP'90), pages 3--17. The MIT Press, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the cost of searching intended programs. For example, few instances and program schema in which variable occurrences are missing might be su#cient for programming. 1. 1 Concurrent Logic Programming The mechanism of error correction in Kima is based on the mode and type systems of Moded Flat GHC [35, 36]. Moded Flat GHC is a concurrent logic (and consequently, a concurrent constraint) language with a constraintbased mode system. Concurrent logic languages [32] provide a simple and powerful model of concurrency as well as a full fledged programming language with . first class message channels, ....

K. Ueda and M. Morita. A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming (ICLP'90), pages 3--17. The MIT Press, 1990.

....call has also weight 1. These weights are assigned to the clauses at compile time by annotating the Prolog program. The scheduler then uses this information in order to select tasks to spawn. This technique was originally used by Tick [34] to control the scheduling of and parallel goals in FGHC [35], but it was used by Wai Keong to control the spawning of or parallel alternatives. Another strategy technique is reported in [21] where a method to remove structural imbalance of the programs by global analysis (basically unfolding flattening recursive predicates) is proposed. By removing ....

Kazunori Ueda and Masao Morita. A New Implementation Technique for Flat GHC. In Proceedings of the Seventh International Conference on Logic Programming, pages 3--17. MIT Press, June 1990.

....programming by restricting its descriptive power somewhat and doing aggressive optimization based on static program analysis. The crucial point here is not to do so in an ad hoc manner. We have thus far developed a constraint based mode system for Flat GHC that enables static dataflow analysis [6, 10]. The resulting language is called Moded Flat GHC. This approach is similar in essence to static type systems found in many languages, and is considered one of the most systematic approaches to syntactic restriction and static analysis. The following observations have been obtained from several ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming, The MIT Press, Cambridge, Mass, June 1990, pp. 3--17.

....name freeze 2. freeze(Var, Goal) calls Goal if Var is instantiated, otherwise Goal is suspended and woken only after Var is instantiated. Similar mechanisms were available in other systems: MU Prolog [9] IC Prolog [4] ESP [2] SICStus Prolog [1] NU Prolog [10] or in committed choice languages [12, 13, 5, 14]. During our designs we have identified the main features which determine particular approaches and answer the important questions. Below we list the main questions that have to be answered by every design. 2.1 When to Delay a Call Which condition has to be satisfied or violated for a call to be ....

Kazunori Ueda and Masao Morita. A new implementation technique for flat GHC. In Proceedings of the 7th ICLP, pages 3--17, Jerusalem, June 1990. 23

....that exploit and parallelism, or parallelism and and or parallelism. Compile time granularity analysis consists of evaluating sizes of goals and clauses and amount of potential parallelism. Research has been done in compile time granularity analysis for concurrent logic languages such as FGHC [21] and for parallel logic programming systems that exploit independent [22] and dependent [28] and parallelism. Besides, Debray et al. have developed the CASLOG system to perform complexity analysis of goals and clauses [4] 5] In this work we present a more complete framework for compile time ....

K. Ueda and M. Morita, A New Implementation Technique for Flat GHC, Proceedings of the Seventh International Conference on Logic Programming (June 1990) 3-17.

....is a similar one of grain size tuning, but under much more difficult circumstances where little synchronization and data reference information may be available at compile time. Concurrent logic programming approaches, particularly those based on the committed choice The Concert System 18 model [40, 17, 33, 32] are similar to concurrent object oriented languages. However, they have little support for encapsulation and parallel collections. In committed choice languages, the emphasis is on task parallelism which is often expressed as operations on a stream. While some degree of grain size tuning can be ....

K. Ueda and M Morita. A new implementation technique for flat GHC. In Proceedings Seventh International Conference on Logic Programming, pages 3--17. MIT Press, 1990. Revised version to appear in New Generation Computing.

....it was able to determine the modes completely when applied to a variety of GHC programs available in the literature (a.o. 23] including append, merge, split, permutation, quick sort and merge sort. The above procedure is simple compared to the procedure presented in Ueda and Morita [30]. This due to the simplicity of our mode system (giving moding only to the arguments of 40 predicates but not to the subterms of the arguments) over that of [30] The complex mode system in [30] was motivated (and justified) by the intended application: optimization of interprocess ....

....split, permutation, quick sort and merge sort. The above procedure is simple compared to the procedure presented in Ueda and Morita [30] This due to the simplicity of our mode system (giving moding only to the arguments of 40 predicates but not to the subterms of the arguments) over that of [30]. The complex mode system in [30] was motivated (and justified) by the intended application: optimization of interprocess communication. Our mode system is a subsystem of theirs and is adequate for transforming GHC programs to term rewriting systems. The procedure presented in [30] derives the ....

[Article contains additional citation context not shown here]

K. Ueda and M. Morita (1990), A new implementation technique for flat GHC, Proc. of Intl. Conf. Logic Prog., ICLP'90, pp. 3-17. Revised version appears as Moded Flat GHC and its Message-oriented Implementation Technique, New Generation Computing 13, pp. 3-43.

....through the evaluation of guards, usually implemented as calls to built in predicates that make clauses mutually exclusive. Once a guard succeeds, the system commits to the clause containing the guard, and all other clauses are ignored. Other examples of these systems include GHC [93, 91] FGHC [94], Concurrent Prolog [83] and KL1 [59] Concurrent Prolog is a logic programming language in that a program is a collection of universally quantified Horn clause axioms, and a computation is an attempt to prove a goal an existentially quantified conjunctive statement from the axioms of the ....

K. UEDA and M. MORITA. A New Implementation Technique for Flat GHC. In Proceedings of the Seventh International Conference on Logic Programming, pages 3--17. MIT Press, June 1990.

....which makes garbage collection algorithms more complex. Some important contributions are by Ciepielewski and Haridi [21] and Warren [52] More recent contributions are [11, 23, 25, 27, 53] Another contribution concerning concurrent logic programming with flat guarded Horn clauses is found in [44]. Parallel or real time implementation of garbage collectors for logic programming is not presented. A study on this topic can, for example, be found in [10] Real time behavior was also one of the incentives in [37] to introduce some kind of generational garbage collector. Another issue left ....

K. Ueda and M. Morita. A new implementation technique for flat GHC. In D.H.D. Warren and P. Szeredi, editors, 7th Int. Conf. Logic Programming, pages 3--17, MIT Press, Jerusalem, 1990.

....the typical case seems to be that most of these processes are suspended during execution, thus being essentially useless for the purpose of parallel speedup. Suspension is so common in concurrent logic langauges that recent implementations optimize this situation rather than computation [73]. Initially, it was thought that compilers could handle this 6 Introduction complexity automatically. Some efforts in this direction have appeared recently [39, 47] though the absolute benefits are unclear. We note that Shen has found that sequential Prolog programs seldom or never exhibit ....

K. Ueda, M. Morita, A new implementation technique for flat GHC, in Logic Programming: Proceedings of the Seventh International Conference, MIT Press, 1990. f5g

....significant work by some sort of scheduler process thus it is entirely possible for the scheduler to become the limiting factor in execution performance. This thesis suggests a two step approach to alleviating these problems, along lines explored earlier by others including Ueda and Morita [30, 31], Tick and Korsloot [19] and King and Soper [15, 17, 16] First, the program is statically analyzed to discover where variables will be bound. This analysis allows traditional compiler optimizations to be applied to the compilation of FGHC programs, and also provides a framework for a second ....

....programmer error in declaring modes, and can lead (as in the algorithm presented here) to a much richer set of derived information. For example we may wish to derive the mode of each variable in the syntactic program, or perhaps modes of objects that are not explicitly mentioned. Ueda and Morita [30] outlined a mode analysis technique that showed great elegance and potential for efficiently deriving rich mode information. Essentially they suggested propagating a set of mode constraints around a control flow graph representing the program. The analysis was restricted to moded flat ....

[Article contains additional citation context not shown here]

K. Ueda and M. Morita. A New Implementation Technique for Flat GHC. In International Conference on Logic Programming, pages 3--17. Jerusalem, MIT Press, June 1990.

....through the evaluation of guards, usually implemented as calls to built in predicates that make clauses mutually exclusive. Once a guard succeeds, the system commits to the clause containing the guard, and all other clauses are ignored. Other examples of these systems include GHC [126] FGHC [127], ConcurrentProlog [128, 74] and KL1 [123] Concurrent Prolog is a logic programming language in that a program is a collection of universally quantifi ed Horn clause axioms, and a computation is an attempt to prove a goal an existentially quantifi ed conjunctive statement from the axioms of ....

Kazunori Ueda and Masao Morita. A New Implementation Technique for Flat GHC. In Proceedings of the Seventh International Conferenceon Logic Programming, pages 3-- 17. MITPress, June 1990.

....arguments is necessary for this optimization to be carried out. For languages such as Parlog [6] this information is available from user annotations. For other logic programming languages, this information may be obtained via dataflow analysis (e.g. see [8] for mode analysis of Prolog programs, [21, 23] for mode analysis of FGHC programs) However, a number of other issues must also be addressed in order to carry out the kind of interprocedural register allocation discussed in the fact 2 example above, including how the merits of alternative output placements should be compared, and how output ....

K. Ueda and M. Morita, "A New Implementation Technique for Flat GHC", Proc. Seventh International Conference on Logic Programming, Jerusalem, June 1990, pp. 3-17. MIT Press.

....sequentialize the program. Finally, threaded code is generated based on the global information. We benefit from restricting ourselves to subset of the CLP family that allows full mode information to be derived. Our choice is Fully Moded Flat Guarded Horn Clauses (FGHC) defined by Ueda and Morita [19], representing a class of languages such as Strand [5] and Janus [14] We call this language FM for short. The essential idea of FM is that the input output structure of corresponding clause arguments of a procedure is required to be identical. Thus, for example, the procedure p(a(X) b(Y ) ....

....over to CLP languages, especially given the extremely fine grained parallelism of traditional implementations. As we have reviewed throughout the paper, work related to the process of compiling for Diadora includes the work of King and Soper [8] on producer consumer analysis, Ueda and Morita s [19] work on mode analysis, and the work of Traub and Culler [18] on forming threads. All of these are, however, conservative, in the interests of safety for traditional implementations. Thus they form smaller threads than might be produced given the deadlock breaking ability of Diadora lazy task ....

K. Ueda and M. Morita. A New Implementation Technique for Flat GHC. In International Conference on Logic Programming, pages 3--17. Jerusalem, MIT Press, June 1990.

....R) In the AKL, the tree can be represented by a tree of processes, as in GHC. It can be argued that this is less efficient than the Prolog solution, but new results on compilation techniques for FGHC like programs suggest that this inefficiency is not necessarily an inherent problem (see [12]) A more definitive solution, allowing fully automatic translation, is to add noisy cut as a new guard operator in the language [8, 10] This is an option in the Kernel Andorra Prolog framework, and it has been included in the prototype implementation of AKL. Noisy cut has the disadvantage of ....

Kazunori Ueda. A new implementation technique for flat GHC. In Proceedings of the Seventh International Conference on Logic Programming. MIT Press, 1990.

.... Objects based on Ports If messages are consumed one at a time by the object message handler, and the input stream is not manipulated in other ways, as can be guaranteed by an object oriented linguistic extension, then it is possible to compile the message handler using message oriented scheduling [19]. Instead of letting messages take the indirect route through the stream, this path can be shortcut by letting the message handling process pose as a special kind of port, which can consume its messages directly. There is then no need to save messages to preserve stream semantics. It is also easy ....

Kazunori Ueda and Masao Morita. A new implementation technique for flat GHC. In Proceedings of the Seventh International Conference on Logic Programming. MIT Press, 1990.

....computation rules. However, it also enables more global analysis which is necessary for ensuring termination. Control information in the form of modes can be compiled into reasonably efficient local tests [Som90] Alternatively, mode information can be inferred from local suspension conditions [UM90] for the purpose of analysing termination, though some precision may be lost in the process. The most flexible way to express modes of logic programs seems to be to attach mode information to nodes of trees which describe the types of the arguments of a procedure (see [Som87] UM90] BLR92] ....

....conditions [UM90] for the purpose of analysing termination, though some precision may be lost in the process. The most flexible way to express modes of logic programs seems to be to attach mode information to nodes of trees which describe the types of the arguments of a procedure (see [Som87] [UM90] [BLR92] Red92] for example) To avoid details of how types and modes are actually defined (or inferred) we simply assume that there is a subset of the Herbrand base which defines the set of well typed atoms. It is assumed that all atoms used in all derivations have a well typed instance. Each ....

Kazunori Ueda and Masao Morita. A new implementation technique for flat GHC. In Proceedings of the Seventh International Conference on Logic Programming, Jerusalem, Israel, June 1990.

....is an impossibly difficult problem. Message passing mechanisms defining import export tables and protocols were developed [40] but little empirical analysis has been published. Compilation techniques to determine runtime characteristics of logical variables, such as hookedness and modes [83], and exploit them to speedup bindings, minimize suspensions, and minimize memory consumption, have not yet been implemented in the current PIM compilers. The ICOT research schedule began with the development of the personal inference machines (PSI I,II,III) followed by mockup PIMs ....

K. Ueda and M. Morita. A New Implementation Technique for Flat GHC. In International Conference on Logic Programming, pages 3--17. Jerusalem, MIT Press, June 1990.

....This paper proposes a framework of automated debugging of program errors under static, constraint based systems for program analysis, and shows how and why program errors can be fixed in the absence of programmers declarations. The language we are particularly interested in is Moded Flat GHC [7][8] proposed in 1990. Moded Flat GHC is a concurrent logic (and consequently, a concurrent constraint) language with a constraint based mode system designed by one of the authors, where modes prescribe the information flow that may be caused by the execution of a program. Supported in part by ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming (ICLP'90), The MIT Press, 1990, pp. 3--17.

....similar to division by zero; for instance, a Pascal statement X : 5 div 0 can be viewed as insisting that different values, X 0 ( 0) and 5, should become identical. 6 Mode System for Flat GHC To guarantee the security of unification statically, a mode system for Flat GHC was designed in 1990 [12]. This and subsequent sections will introduce it rather informally but from various aspects. We refer the readers to [15] for the precise and detailed description. As a metaphor, consider two electric devices connected by a cable for conveying signals. The cable may have a structure such as an ....

....The reference counting worked quite efficiently on Parallel Inference Machines, but a static scheme is clearly more desirable on stock microprocessors. Analysis of the mode and the number of receivers enables a sophisticated implementation technique that we call message oriented implementation [12] [15] Message oriented implementation compiles unification for stream communication into low latency message passing, which is particularly beneficial to programs in which processes suspend frequently and incur much process switching overhead. 8.3 Language Design We are currently exploring the ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming, The MIT Press, Cambridge, MA, 1990, pp. 3--17.

....goal. This alternative, called eventual publication, is advantageous for implementation, though some programming techniques can be used only in atomic publication languages. Interestingly, our choice of eventual publication recently lead to the idea of the messageoriented scheduling of goals [25], a scheduling that contrasts sharply with the ordinary one. Moreover, GHC enjoys anti substitutivity [21] a property which allows the delay of interprocess communication between two occurrences of a shared variable. Anti substitutivity allows two occurrences of the same variable to have even ....

....is that two or more processes sharing a variable may try to instantiate it non cooperatively. Doc [5] and Janus [17] introduced annotations (attached to occurrences of variables) to syntactically guarantee that only one process can instantiate a variable. On the other hand, Ueda and Morita [25] showed that simple mode analysis can be used to guarantee the same property. The mode system provides a uni ed framework for mode declaration (of which annotation is one possible way) mode inference and mode checking. Restriction to one producer per variable disallows a variable to be used as a ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming, MIT Press, 1990, pp. 3-17. { 9 {

....techniques that optimize the latency rather than the throughput of interprocess communication are badly needed for executing those programs efficiently. This is another major topic of the paper, which will be discussed in Section 3. This paper is a revised and extended version of an earlier paper [22]. In particular, the mode system first proposed in [22] is described here in much more detail, with proofs of important properties and a study of the cost of mode analysis. 2. Moded Flat GHC The first step towards the optimization of interprocess communication is to analyze what forms of ....

....of interprocess communication are badly needed for executing those programs efficiently. This is another major topic of the paper, which will be discussed in Section 3. This paper is a revised and extended version of an earlier paper [22] In particular, the mode system first proposed in [22] is described here in much more detail, with proofs of important properties and a study of the cost of mode analysis. 2. Moded Flat GHC The first step towards the optimization of interprocess communication is to analyze what forms of communication will take place when a program is executed. This ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming, Warren, D. H. D. and Szeredi, P. (eds.), MIT Press, 1990, pp. 3--17.

....Implementation of Moded Flat GHC Kazunori Ueda NEC C C Systems Research Laboratories 1 1, Miyazaki 4 chome, Miyamae ku, Kawasaki 213, Japan ueda csl.cl.nec.co. jp Masao Morita Mitsubishi Research Institute 3 6, Otemachi 2 chome, Chiyoda ku, Tokyo 100, Japan Abstract We proposed in [Ueda and Morita 1990] a new, message oriented implementation technique for Moded Flat GHC that compiled unification for data transfer into message passing. The technique was based on constraint based program analysis, and significantly improved the performance of programs that used goals and streams to implement ....

....issue which is outside the scope of this paper. Processes used as storage are almost always suspending, but should respond quickly when messages are sent. However, most implementations of concurrent logic languages have not been tuned for processes with this characteristic. In our earlier paper [Ueda and Morita 1990], we proposed message oriented scheduling of goals for sequential implementation, which optimizes goals that suspend and resume frequently. Although our primary goal was to optimize storage intensive (or more generally, demand driven) programs, the proposed technique worked quite well also for ....

[Article contains additional citation context not shown here]

K. Ueda and M. Morita, A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming, MIT Press, 1990, pp. 3--17. A revised, extended version to appear in New Generation Computing.

....The framework was then implemented in Kima, an automated error correction system for concurrent logic programs, which featured several improvements to make the system more practical and efficient. The mechanism of error correction in Kima is based on the mode and type system of Moded Flat GHC [8][9] Moded Flat GHC is a concurrent logic (and consequently, a concurrent constraint) language with a constraint based mode system designed by one of the authors. Languages equipped with strong typ 2 ing or strong moding 1 enable the detection of type mode errors by checking or reconstructing ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming (ICLP'90), The MIT Press, 1990, pp. 3--17.

....The framework was then implemented in Kima, an automated error correction system for concurrent logic programs, which featured several improvements to make the system more practical and e#cient. The mechanism of error correction in Kima is based on the mode and type system of Moded Flat GHC [8][9] Moded Flat GHC is a concurrent logic (and consequently, a concurrent constraint) language with a constraint based mode system designed by one of the authors. Languages equipped with strong typing or strong moding 1 enable the detection of type mode errors by checking or 1 Modes can be ....

Ueda, K. and Morita, M., A New Implementation Technique for Flat GHC. In Proc. Seventh Int. Conf. on Logic Programming (ICLP'90), The MIT Press, 1990, pp. 3--17.

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