S.K. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, pages 48--62, Cambridge, MA, June 1997. MIT Press, Cambridge, MA.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....is of crucial importance. Nevertheless, relations which are not supposed to fail are quite common. We say that a relation is not supposed to fail if when called in a correct way produces at least one answer (e.g. sort, flatten and append) The ubiquity of such predicates is con rmed by [4, 11, 5] and the fact that Mercury [14] employs a mode system which actually subsume our partitioning. A partitioning is a map from the set of predicate symbols into the set ftest; non testg. We also say that the program P is correctly partitioned wrt the query A i every time that a non test atom B is ....

....are test. Using a non trivial partitioning has the advantage of undercovering the lazy aspects of logic programming, which is what we want to do. Checking correctness is orthogonal to the purposes of this paper, but we should mention that it can be done either using abstract interpretation [5] or on modes and types [4, 11] The Translation Clearly, we need to capture the fact that a predicate might succeed (returning a computed answer) or fail. To do so, we introduce the datatype Result in our Haskell programs by: data Result = Suc j Fail That is, the datatype Result has two ....

S. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In Lee Naish, editor, Proceedings of the 14th International Conference on Logic Programming, pages 48-62, Cambridge, 1997. MIT Press.

....p with types in T . When p is defined also by facts, the introduction of the dummy test true is necessary for satisfying Definition 6. Verifying exhaustiveness of tests for a given predicate and input types is not easy, it is a semantic condition and not decidable in general, as shown in [18]. In the following Section we will discuss this issue by showing cases in which the verification is simple and we will propose simplifications which can help in the verification. For the moment we assume to be able to verify the exhaustive property when necessary. Since we assume that we can ....

....prove that they are a set of exhaustive tests for that predicate. Atoms which are eligible as tests are the ones which do not satisfy noFD conditions (condition 2 in Definition 8) Proving that they are exhaustive is not trivial. It is a semantic condition and not decidable in general, as shown in [18], but in most practical cases the conditions in Definition 8 can be verified. Let us consider Example 3. odd and even are non noFD predicates, hence also oddLength and evenLength which depend on them are non noFD. In the definition of posP ivot we choose them as tests. We should prove that ....

[Article contains additional citation context not shown here]

Debray, S., Lopez-Garcia, P., Hermenegildo, M.: Non-Failure Analysis for Logic Programs. In Naish, L. (ed.): Proceedings of the International Symposium on Logic Programming, The MIT Press (1997) 48--62

....obtained, using the approach described above. ii) we can use abstract interpretation tools built for logic programming to abstract the concrete denotation and use that for verifying the properties; in fact, work is in progress in our group to use non failure analysis of constraint logic programs [4] to verify properties of real time systems (see also Section 7.2) 5 Specification, Implementation and Verification of DSL Programs In this section we discuss how our logical, denotational approach can be applied to specification, implementation, and verification of software systems, especially ....

.... vari(V) hats(H) term(T) op1(op(H, hats(H) op1(op( op1(op( hats( hats( L] hats(L) vari(x) x] vari(y) y] vari(z) z] digit(0) 0] digit(1) 1] digit(2) 2] digit(3) 3] digit(4) [4]. digit(5) 5] digit(6) 6] digit(7) 7] digit(8) 8] digit(9) 9] SEMANTICS: sexp(e(X) INL,L) sterm(X,INL,ONL,L1) getlist(ONL, Bl) append(L1,Bl,L) sexp(e(T,O,E) NL,L) sterm(T,NL,OL, L1) sop(O,OL,NL1) CloseBraces is OL NL1, getlist(CloseBraces, Bl) ....

[Article contains additional citation context not shown here]

S. Debray, P. Lopez-Garcia, and M. Hermenegildo. Non-failure Analysis for Logic Programs. In International Conference on Logic Programming. MIT Press, 1997.

....included. Proposals have been made to attempt improving the execution behaviour in presence of potentially nondeterministic and parallel computations. Hermenegildo has presented various static analysis schemes whose results can be applied to improve management of and parallel executions (e.g. [8]) Hermenegildo and Carro [23] described also an optimization scheme, called backtracking families which allows to take advantage of special features of the computation to improve backtracking. Related optimizations have been presented by Gupta and Pontelli [12] In the context of dependent ....

....parallel calls. The performance results achieved are remarkable. We believe the performance of the system can be further improved by making a more intense use of static analysis. In particular, we are currently exploring the possibility of using non failure analysis and determinacy analysis [22, 8] to reorganize the parallel calls and drive the backtracking process in the best possible way (e.g. actually reusing side effect free deterministic goals) We are also studying how to improve the performance of backward execution in the context of DAP, following the initial ideas by Shen [26] ....

Debray, S., Lopez-Garcia, P., and Hermenegildo, M. Non-failure Analysis for Logic Programs. In Int. Conference on Logic Programming (1997), MIT Press.

....denotation obtained, using the approach described above. ii) we can use abstract interpretation tools built for logic programming to abstract the concrete denotation and use that for verifying the properties; in fact, work is in progress in to use non failure analysis of constraint logic programs [8, 10] to verify properties of real time systems. A third approach that can be used to remove the niteness restriction is to use rst order theorem proving. The logical denotation of a program provides an axiomatization w.r.t. the language s semantics. These axioms can then be fed to a theorem prover, ....

S. Debray, P. Lopez-Garcia, and M. Hermenegildo. Non-failure Analysis for Logic Programs. In International Conference on Logic Programming. MIT Press, 1997.

....precise understanding of strengths and limitations of the proposed approach. The kind of parallelism addressed by the method is, intuitively, that of dynamic (nested) pipeline networks. Another, syntactic, relation 2 is proposed, for which an automatic check method has been recently proposed in [9]. In conclusion, we discuss some examples describing applicability and limitations of the approach, that open natural research directions. We also discuss how, in its present form, the proposed method represents a strict improvement on the existing approaches for termination analysis of guarded ....

....is a variable, are an acceptable program and query. Thus, we are in 12 the hypotheses of Theorem 16, which implies that every SLDG derivation for Merge and omerge(In 1 , In 2 , Out) via any safe selection rule is successful. 6 Discussion 6. 1 Covering algorithm The recent work by Debray et al. [9] helps in proving that a program is failure free for a given . They provide a method that, given mode and type information, can detect whether the clauses defining a predicate cover the type of the predicate. In particular, let us consider relation 2 . The non failure analysis of [9] is based ....

[Article contains additional citation context not shown here]

S. Debray, P. L'opez-Garc ' ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the 1997 International Conference on Logic Programming. The MIT Press, 1997.

....A in P then B has at least one successful LD derivation. Thus every program is correct wrt. the trivial partitioning in which all predicates are test. Checking correctness is orthogonal to the purposes of this paper, but we should mention that it can be done either using abstract interpretation [DLGH97] or on modes and types [PR97] also Mercury employs a system based on modes and types in order to check that the programs are consistent (modulo non termination) wrt. the partitioning provided by the programmer. The partition into test and non test predicates exposes the implicit failure ....

S. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In Lee Naish, editor, Proceedings of the 14th International Conference on Logic Programming, pages 48-62, Cambridge, 1997. MIT Press. 18

....proof obligations of the method. In particular, a weakest persistent relation 0 is characterized, which gives us a precise understanding of strengths and limitations of the proposed approach. Another, syntactic, relation 2 is proposed, for which an automatic check method has been proposed in [9]. Our object of study is an extension of the logic programming paradigm obtained by combining guarded clauses and delay declarations . The former are taken from parallel logic programming languages [20, 22] whilst the latter were introduced in MU Prolog [17] and further adopted by many logic ....

....hand, we provide a general framework for proving failure freedom and termination with respect to any safe SLDG selection rule. Special cases of guarded programs with delay declarations include de nite logic programs, parallel logic programs, and G odel programs. Covering algorithm Debray et al. [9] provide a method that, given mode and type information, can detect whether the clauses de ning a predicate cover the type of the predicate. We claim that their approach helps us in proving that a program is failure free for a given . In particular, let us consider relation 2 . The ....

[Article contains additional citation context not shown here]

S. Debray, P. Lopez-Garc  ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the 1997 International Conference on Logic Programming. The MIT Press, 1997.

....et al. 2] extends functionality analysis by estimating the number of answers to queries (instead of whether is more or less than 1) The determinacy analysis implemented in Mercury compiler [10] also estimates the cardinality of each query. Non failure analysis proposed by Debray and Hermenegildo [7] determines the set of goals that cannot fail: i.e. have at least one answer. Mutual exclusion analysis described by Post [18] infers predicates with the property that at most one clause becomes applicable at clause selection time. All the above methods attempt to classify the predicates in a ....

S. Debray, P. Lopez-Garcia, and M. Hermenegildo. Non-failure analysis of logic programs. In ICLP, pages 48--62, 1997.

....al. 2] extends functionality analysis by estimating the number of answers to queries (instead of whether is more or less than 1) The determinacy analysis implemented in Mercury compiler [10] also estimates the cardinality of each query. Non failure analysis proposed by Debray and Hermenegildo [7] determines the set of goals that cannot fail: i.e. have at least one answer. Mutual exclusion analysis described by Post [18] infers predicates with the property that at most one clause becomes applicable at clause selection time. All the above methods attempt to classify the predicates in a ....

S. Debray, P. Lopez-Garcia, and M. Hermenegildo. Nonfailure analysis of logic programs. In ICLP, pages 48--62, 1997.

....a length of less than four. Put another way, if the length of argument is known to be greater or equal to four, then the computation must either exceed dmin resolution steps or fail. Possible failure (or equivalently definite non failure) can be detected with a query dependent non failure analysis [8]. Thus if the program is queried with a Qs 2 goal where the first argument is known to be a list of integers, say, then non failure can be deduced [8] Hence, if the argument is also known to have a length of greater or equal to four, then the goal is guaranteed to lead to a computation that ....

....dmin resolution steps or fail. Possible failure (or equivalently definite non failure) can be detected with a query dependent non failure analysis [8] Thus if the program is queried with a Qs 2 goal where the first argument is known to be a list of integers, say, then non failure can be deduced [8]. Hence, if the argument is also known to have a length of greater or equal to four, then the goal is guaranteed to lead to a computation that exceeds dmin resolution steps. 3 Preliminaries Syntax of logic programs Let Func, P red and V ar respectively denote the set of function symbols, ....

[Article contains additional citation context not shown here]

S. Debray, P. Lopez Garca, and M. Hermenegildo. Non-Failure Analysis of Logic Programs. In ICLP'97. MIT Press, 1997.

....In program verification and transformation we are actually interested in identifying both the queries which have some successful LD derivations and those which have some finitely failing LD derivations. Non failure analysis is interesting also for parallel execution optimization. In [14] a method is given for detecting programs and queries which produce at least one solution or do not terminate. The method is based on a different notion of mode and type information and deals also with programs containing tests. Our proposal is clearly more restrictive, but also simpler to ....

S. Debray, P. Lopez-Garcia, and M.Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the International Symposium on Logic Programming, pages 48--62, 1997.

....In program verification and transformation we are actually interested in identifying both the queries which have some successful LD derivations and those which have some finitely failing LD derivations. Non failure analysis is interesting also for parallel execution optimization. In [17] a method is given for detecting programs and queries which produce at least one solution or do not terminate. The method is based on a different notion of mode and type information and deals also with programs containing tests. Our proposal is clearly more restrictive, but also simpler to ....

S. Debray, P. Lopez-Garcia, and M.Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the International Symposium on Logic Programming, pages 48--62, 1997.

....prove that they are a set of exhaustive tests for that predicate. Atoms which are eligible as tests are the ones which do not satisfy noFD conditions (condition 2 in Definition 8) Proving that they are exhaustive is not trivial. It is a semantic condition and not decidable in general, as shown in [18], but in most practical cases the conditions in Definition 8 can be verified. Let us consider Example 3. odd and even are non noFD predicates, hence also oddLength and evenLength which depend on them are non noFD. In the definition of posP ivot we choose them as tests. We should prove that ....

....analysis. Our work is clearly related to those on cardinality analysis, for example in [20, 17] In these works an approximation of the number of c.a.s. for a query is estimated by means of Abstract Interpretation. Success analysis is interesting also for parallel execution optimization. In [18] a technique is proposed for detecting programs and queries which produce at least one solution or do not terminate. The technique deals with programs containing tests and it is based on a different notion of mode and type information from ours. An algorithm is given for unification and arithmetic ....

[Article contains additional citation context not shown here]

S. Debray, P. Lopez-Garcia, and M.Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the International Symposium on Logic Programming, pages 48--62, 1997.

.... existing Prolog systems, Mercury has integrated determinacy analysis into its compiler [9] This notion of determinacy analysis, where again, the number of answers of a user defined predicate is of importance, also crops up in works of Debray and Hermenegildo in the domain of non failure analysis [6]. Our method fundamentally differs from these methods in the fact that rather than creating a classification of the program predicates (such as functional nonfunctional) we compute the necessary condition of every program clause and then use them selectively to promote early failure determinacy. ....

S. Debray, P. Lopez-Garcia, and M. Hermenegildo. Non-failure analysis of logic programs. In Proceedings of 1997 ICLP, pages 48--62, 1997.

No context found.

S.K. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, pages 48--62, Cambridge, MA, June 1997. MIT Press, Cambridge, MA.

No context found.

S.K. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, pages 48--62, Cambridge, MA, June 1997. MIT Press, Cambridge, MA.

No context found.

S. Debray, P. Lopez Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997.

....a type, that is efficient in practice. Based on this information, we show how to identify goals and procedures that can be guaranteed to not fail at runtime. Space limitations prevent us from discussing several issues completely or including some of the longer proofs. We refer the reader to [4] for details. 2 Preliminaries We assume an acquaintance with the basic notions of logic programming. In order to reason about non failure, it is necessary to distinguish between unification operations that act as tests (and which may fail) and output unifications that act as assignments (and ....

....Then empty(M; S) terminates, and if it returns true then M Omega S . For each of these theorems, correctness can be argued by induction on the depth of recursion of functions empty1 and empty2 upon termination; the termination arguments follow standard lines. Complete proofs can be found in [4]. One reason for imprecision in the case of non tuple distributive regular types is that the function intersection(M; A) described above computes a superset of the exact intersection when we deal with general regular types (this result can be derived from the work of Dart and Zobel [3] Another ....

S. Debray, P. L'opez Garc'ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. Technical Report TR Number CLIP20/96.0, T.U. of Madrid (UPM), Facultad Inform'atica UPM, 28660-Boadilla del Monte, Madrid-Spain, November 1996.

....is a superset of the set of argument values that are actually encountered at runtime. Unfortunately, straightforward attempts to address this issue, for example by trying to infer lower approximations to the calling types of predicates, fail to yield nontrivial lower bounds for most cases. In [2], we showed how, given mode and (upper approximation) type information, we can detect procedures and goals that can be guaranteed to not fail. Our technique is based on an intuitively very simple notion, that of a (set of) tests covering the type of a variable. We showed that the problem of ....

S. Debray, P. L'opez Garc'ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, Leuven, Belgium, June 1997. MIT Press, Cambridge, MA.

....the domains for the combination [CC79, CMB 95] Two other domains, a modified version of Path sharing [KS95] and Aeqns (abstract equations) MSJB95] are currently being incorporated to the system. In addition the system performs type and determinacy non failure analysis [BJ88, GdW94, DLH97] 3.2.2 Constraint Programming: Several domains are available, some of which have been implemented by other users of the PLAI system, notably 2 Some of these domains have been implemented by other users of the PLAI system, notably the K. U. Leuven, Monash University, and the U. of Melbourne. ....

S. K. Debray, P. L'opez Garc'ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, pages 48--62, Leuven, Belgium, June 1997. MIT Press, Cambridge, MA. 15

....that a goal has a single solution. In addition, knowledge that a goal will not fail is also quite useful [27] It is beyond the scope of this paper to address how this information is gathered the reader is referred to related work in the area of abstract interpretation based global analysis [33, 4, 11, 44, 14]. We will address instead how such information can be exploited at the parallel abstract machine level. We start by considering the case in which several parallel goals, perhaps not contiguous in the program text, but which are known to be deterministic, end up being executed on the same ....

.... Goals Prolog Execution Executed unoptimized optimized deriv(1) 261 [9] deriv(2) 2109 [11] deriv(3) 16893 [11] deriv(4) 135165 [11] boyer(0) 24 [3 8] boyer(1) 747 [111 153] boyer(2) 7290 [543 745] boyer(3) 282168 [4770 6500] quick sort(50) 150 [13 15] quick sort(100) 300 [14 16] quick sort(150) 450 [15 17] quick sort(200) 600 [15 16] poccur(1) 30 [9 10] poccur(2) 60 [11 12] poccur(3) 90 [11 13] poccur(4) 120 [12 13] poccur(5) 150 [12 13] takeuchi(13) 1412 [19 23] takeuchi(14) 4744 [21 29] takeuchi(15) 10736 [21 30] takeuchi(16) ....

S. Debray, P. L'opez-Garc'ia, and M. Hermenegildo. Non-failure Analysis for Logic Programs. Technical Report CLIP14/94.0, Univ. Politecnica de Madrid, October 1994.

.... list(A,num) num(B) term(C) term(D) list(A,num) num(B) list(C,num) list(D,num) true pred append(A,B,C) list(A,num) list1(B,num) term(C) list(A,num) list1(B,num) list1(C,num) Non failure and Determinacy Analysis: Ciaopp includes a non failure analysis, based on [6], which can detect procedures and goals that can be guaranteed not to fail, i.e. to produce at least one solution or not terminate. It also can detect predicates that are covered , i.e. such that for any input (included in the calling type of the predicate) there is at least one clause whose ....

S. K. Debray, P. Lopez Garca, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, pages 48-62, Cambridge, MA, June 1997. MIT Press.

.... in recursive data structures [8,29,34,36] These analyses have been applied to the detection of both strict and non strict independence [8,9] Analyses have been developed also to derive other impor tant properties beyond variable instantiation states such as determinism [39] non failure [13], and number of answers [6] Dealing with Irregularity and Speculation Dynamic Solutions: The preceding discussion has on purpose avoided the issue of run time overheads. The obvious practical implication of the existence of overheads (task creation, scheduling, data movement, etc. is that ....

S. K. Debray, P. L'opez Garc'ia, and M. Hermenegildo. Non-Failure Analysis for Logic Programs. In 1997 International Conference on Logic Programming, Leuven, Belgium, June 1997. MIT Press, Cambridge, MA.

No context found.

S. Debray, P. Lopez-Garca, and M. Hermenegildo. Non-failure analysis for logic programs. In Proceedings of the 1997.

No context found.

S. Debray, P. Lopez-Garcia, and M.Hermenegildo. Non-Failure Analysis for Logic Programs. In Proceedings of the International Symposium on Logic Programming, pages 48--62, 1997.

No context found.

) Debray, S., and Lopez-Garcia, P., and Hermenegildo, M. Non-failure Analysis for Logic Programs. In Proc. of the International Conference on Logic Programming, MIT Press, 1997, pp. 48-62.

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