Turchin, V.: 1988, `The Algorithm of generalization in the supercompiler'. In: D. Bj#rner, A. Ershov, and N. Jones #eds.#: Proc. of the IFIP TC2 Workshop on Partial Evaluation and Mixed Computation. pp. 531#549.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Analysis Given an L program pgm and data dat, the semantics of neighborhood analysis is the set of all program data pairs that compute the same result as pgm applied to dat. Neighborhood analysis has, among others, applications to program testing [2] and termination of supercompilation [52], because it tells us how much we can change pgm and dat without a#ecting the output of a computation. Definition 20 (NAN dialect) Let L be a programming language, then L # is a nonstandard NAN dialect of L i# for all pgm DatL, the result of computation (pgm dat # ans) is a pair ans = ....

....passes the KMP test in TSG and in MP. 11.1. Supercompiler: an Equivalence Transformation Modifier We outline an algorithm for supercompilation in TSG which will be our algorithm for equivalence transformation. A more detailed discussion of supercompilation and its methods can be found in [3, 26, 45, 51, 52, 53]) The supercompiler for TSG was implemented in Gofer (330 lines of pretty printed source text) Algorithm 25 [SCP (outline) Given TSG program pgm and class cls, the algorithm starts by driving the initial configuration pgm cls and builds a process tree gr (initially one node labeled with pgm ....

[Article contains additional citation context not shown here]

V. F. Turchin, "The algorithm of generalization in the supercompiler," in [9], pp. 531--549.

....performs folding [9] operations by recursively specializing parts of the program, and building residual invocations of these specializations. These specializations may be invoked from multiple residual call sites, allowing them to be re used. Program specializers also perform generalization [43, 45, 37, 6] operations, ignoring some specialization time information and prohibiting some reductions in order to guarantee that only a finite number of specializations are built. Ideally, a program specializer would perform as many reductions as possible 1 (while still terminating) at specialization ....

....the online parameterized partial evaluation semantics on p. 99 of [15] and the if handler code on p. 11 of [46] The second is when computing the argument values to be used when building a specialization (discussions of generalization can be found in Section 3 of [45] Section 2. 2 of [36] and [43]. In both cases, some loss of information is unavoidable, but can be minimized through the use of a more precise type system which can compute less overly general upper bounds. In this paper, we will show two instances where present systems lose information unnecessarily, along with means for ....

[Article contains additional citation context not shown here]

V. Turchin. The algorithm of generalization in the supercompiler. In D. Bjrner, A. P. Ershov, and N. D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531--549. NorthHolland, 1988.

....0. 6 Related Work Partial Evaluation Partial Evaluation (PE) is not a new invention; the term was coined in 1964 in the context of coping with incomplete information in LISP evaluation [81] First PE like optimizations can be traced back to the REFAL compiler [33] and Turchin s supercompiler [114, 115]; one theoretical milestone was the definition of the Futamura projections [52] cf. section 5.3) Due to its origins in LISP evaluation, most early PE systems were based on the on line principle, i.e. they reconstructed the program during its execution. An example is the REDFUN system [10, 56] ....

V.F. Turchin, The algorithm of generalization in the supercompiler, in: [17], pp. 531-549

....is guaranteed and results are satisfactory This problem has been tackled from two (until now) largely separate angles: the so called off line versus on line approaches. Partial evaluation of functional programs [8, 31] has mainly stressed the former, while supercompilation of functional [67, 68, 65] and partial deduction of logic programs [21, 64, 4, 6, 52, 54] have concentrated on on line control. Some exceptions are [69, 56, 38, 33] Basically, within the off line approach, an analysis phase (by hand and or automatically) prior to the specialisation proper, provides annotations that ....

....prove equally beneficial. Some additional discussions, further motivating the use of global trees and characteristic atoms, can be found in Appendix D. We conclude this section with a brief discussion on the relation between our global control and what may be termed thus in supercompilation [67, 68, 65]. A distinction between local and global control is not yet made in supercompilation, but we feel that this situation is likely to change in the near future. We already pointed out that the inspiration for using derives from [50] and [65] In the latter, a generalisation strategy for positive ....

[Article contains additional citation context not shown here]

V. F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjrner, A. P. Ershov, and N. D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

.... deduction [23, 26] conjunctive partial deduction [16] compiling control [10] loop absorption [31] partial evaluation of functionallogic languages [1] unfold fold transformation of functional programs [11] unfold fold transformation of logic programs [38] tupling [4, 29] supercompilation [39, 40], positive supercompilation [18, 35] generalized partial computation [15] deforestation [42] and online partial evaluation of functional programs [43, 33, 21] Although offline transformers (i.e. transformers making use of analyses prior to the transformation to make changes in the program ....

V.F. Turchin. The algorithm of generalization in the supercompiler. In Bjørner et al. [5], pages 531--549.

....is present in the notion of minimal foldable upper portion of an unfolding tree. Turchin s supercompiler [32, 14] can also do deforestation. Supercompilation performs driving (normal order unfolding and unification based information propagation) and generalisation (a form of abstraction) [33, 30]. Tree structures are used to record the history of configurations [12] The connection between driving and classical partial deduction was established in [13] Recently, a transformation scheme has been proposed for functional logic languages based on an automatic unfolding algorithm that builds ....

V.F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov, N.D. Jones (eds.), Partial Evaluation and Mixed Computation. 531--

....conceived by Turchin (1979,1980,1986) in the early 1970 s in Russia for the language Refal, achieves the effects of both deforestation and partial evaluation, as well as some more dramatic optimizations. This is done by driving, i.e. unfolding and propagation of information, and generalization (Turchin, 1988), a form of abstraction which enables folding. The decision when to generalize is taken online. Generalized partial computation (GPC) due to Futamura (1988) has similar effects and power as supercompilation, but requires the use of a theorem prover. The above methodologies have been developed ....

Turchin, V.F. 1988. The algorithm of generalization in the supercompiler. In Partial Evaluation and Mixed Computation, Bjørner D., Ershov A.P. and Jones N.D. (eds.), pages 341--353, North-Holland.

....Refal and containing a brief discussion on the philosophy underlying the project, and [Tur86b,Tur86c] describing the driving part of the supercomiler in detail. In 1987 Turchin published the paper [Tur87] which gives a foundation of mathematics in terms of metasystem transition. In 1988 the paper [Tur88] described the means of automatically ensuring termination of supercompilation. As far as the author knows, it is still the only paper on automatic termination of the supercompiler. In 1989 the first implementation of a self applicable supercompiler was reported by Gluck and Turchin [Glu89] ....

....pattern formulated in this section. Section 12.5 shows that the problem of deciding for arbitrary term and program whether W terminates is recursively unsolvable. Section 12.6 shows that there are recursive functions for which application of W to any M 1=2 formulation fails to terminate. 1 In [Tur88]. The comment apparently originally comes from a referee report by Olivier Danvy (personal communication with Danvy) 108 CHAPTER 12. INTRODUCTION TO THE PROBLEM OF ENSURING TERMINATION OF W 12.1 The canonical non W termination patterns The problem of the Accumulating Parameter Example 12.1.1 ....

[Article contains additional citation context not shown here]

V. F. Turchin. The Algorithm of Generalization in the Supercompiler. In Partial Evaluation and Mixed Computation. Eds. A. P. Ershov, D. Bjrner, N. D. Jones North-Holland, 1988.

....others, program specialization, program inversion and theorem proving. Other related aspects have been investigated in [1,7,8,13,14,17,18,26] The notion of perfect process graphs and perfect driving were introduced in [22,23] The language S Graph is closely related to Turchin s Refal graphs [25]. But due to SGraph s simpler data structure, untyped variables and only two elementary contractions, one may build rather clear and concise driving algorithms. In particular, there is only one way to compose and decompose S expressions (as opposed to Refal data structures) There is a close ....

Turchin V. F., The algorithm of generalization in the supercompiler. In: Bjørner D., Ershov A. P., Jones N. D. (ed.), Partial Evaluation and Mixed Computation . (Gl. Avernæs, Den - mark). 341-353, 1988.

....contains (in some sense) a term that had previously been seen and in such case inform F that the term be handled with care since it is infinitely growing. Turchin has such a method for determining on line, i.e. during transformation, which terms should be handled with care, for his supercompiler [Tur88]. On line methods have also been devised for partial evaluators [Wei91] We are, however, interested in an off line strategy. Traditionally, selfapplicable partial evaluators have been off line [Bon90] Jon91] Jon92] Online evaluators have also been self applied with some success recently but ....

....A transformation algorithm doing both deforestation and partial evaluation is likely to be similar to Turchin s supercompiler which performs these transformations. Therefore, a method of ensuring termination for the former transformation should be compared to Turchin s on line strategy [Tur88] for his supercompiler. 11.2 Directions for further research First of all the method should be implemented and experimental results for it should be compared with other experimental results. We are only presently aware of the experimental results in [Gil93] The grammar algorithm as we have ....

Valentin F. Turchin. The Algorithm of Generalization in the Supercompiler. In Partial Evaluation and Mixed Computation. Eds. A. P. Ershov, D. Bjrner & N. D. Jones North-Holland, 1988.

....guarantee that the set of or nodes remains finite. Many techniques have been developed for global control of partial evaluation. Such techniques make use of data structures which are very related to the and or analysis graph such as characteristic trees [GB91] Leu95] related to neighborhoods [Tur88]) trace terms [GL96] and global trees [MG95] and combinations of them [LM96] Thus, it seems possible to adapt these techniques to the case of abstract interpretation and formalize them as widening operators. 6 Related Work The integration of partial evaluation and abstract interpretation has ....

V. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov, and N.D. Jones, editors, Proc. of the IFIP TC2 Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

....2: Example derivations 4 Partial deductions using supercompilation As hinted by Figure 2(b) partial evaluation might be used to generate nontrivial derivation tree structure in the presence of unknowns. We have found Gluck and S rensen s formulation [14] of Turchin s supercompilation algorithms [30, 31, 32] useful, because it performs specialization and deforestation simultaneously on line [25, 14, 28] Here is short review of the framework as presented in [14] A source program consists of a set of first order, equationally defined functions and an expression to be simplified. The supercompilation ....

V.F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov, and N.D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

....will guarantee that the set of or nodes remains finite. Many techniques have been developed for global control of partial evaluation. Such techniques make use of data structures which are very related to the and or graph such as characteristic trees [GB91] Leu95] related to neighbourhoods [Tur88]) trace terms [GL96] and global trees [MG95] and combinations of them [LM96] Thus, it seems possible to adapt these techniques to the case of abstract interpretation and formalize them as widening operators. 6 Future Work It remains as future work to experiment with the techniques presented ....

V. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov, and N.D. Jones, editors, Proc. of the IFIP TC2 Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

.... attributes such as types or arithmetic signs [45, 16] Online specializers use a variety of mechanisms to cause the building of specializations: some are based on explicit reduce residualize annotations [20] others on a metalanguage [10] and still others on various forms of recursion detection [45, 43]. The meta language approach allows the user to define a fold here predicate for each function, which is passed the function s arguments and decides whether or not to unfold it. The recursion detection mechanisms compute call histories (essentially a representation of the pending calls in the ....

....general function instance, online specializers achieve generality by modifying the argument values on which the specialization will be built, thus indirectly affecting which reductions will be performed during the construction of the general instance. This process is usually called generalization [43, 45], but has also been referred to as generalized restart [39] After deciding to build a specialization, the specializer finds the set of call sites for which the specialization must be applicable, then computes the least general argument specification which includes the argument specifications of ....

V. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A. P. Ershov, and N. D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531-- 549. North-Holland, 1988.

....the graph. The driving process does not preserve the semantics, as it can extend the domain of functions (as noted by Gluck and S rensen [GS94] Jones et al. JGS93] and S rensen et al. SGJ94] Techniques to ensure termination of driving are studied in S rensen and Gluck [SG95] and Turchin [Tur88] The idea of Turchin [Tur88] is to supervise the construction of the tree and, at certain moments, loop back, i.e. fold a configuration to one of the previous states, and in this way construct a finite graph. The generalization operation which makes it possible to loop back the current ....

.... does not preserve the semantics, as it can extend the domain of functions (as noted by Gluck and S rensen [GS94] Jones et al. JGS93] and S rensen et al. SGJ94] Techniques to ensure termination of driving are studied in S rensen and Gluck [SG95] and Turchin [Tur88] The idea of Turchin [Tur88] is to supervise the construction of the tree and, at certain moments, loop back, i.e. fold a configuration to one of the previous states, and in this way construct a finite graph. The generalization operation which makes it possible to loop back the current configuration is often necessary. In ....

V.F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov, and N.D. Jones, editors, Proceedings of the International Workshop on Partial Evaluation and Mixed Computation, pages 531--549, Amsterdam, 1988. North-Holland.

No context found.

Turchin, V.: 1988, `The Algorithm of generalization in the supercompiler'. In: D. Bj#rner, A. Ershov, and N. Jones #eds.#: Proc. of the IFIP TC2 Workshop on Partial Evaluation and Mixed Computation. pp. 531#549.

No context found.

V.F. Turchin. The Algorithm of Generalization in the Supercompiler. In D. Bjrner, A.P. Ershov, and N.D. Jones, editors, Proc. of the Int'l Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, Amsterdam, 1988.

No context found.

V.F. Turchin. The Algorithm of Generalization in the Supercompiler. In D. Bjrner, A.P. Ershov, and N.D. Jones, editors, Proc. of the Int'l Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, Amsterdam, 1988. This article was processed using the L A T E X macro package with LLNCS style

No context found.

V.F. Turchin. The Algorithm of Generalization in the Supercompiler. In D. Bjrner, A.P. Ershov, and N.D. Jones, editors, Proc. of the Int'l Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, Amsterdam, 1988.

No context found.

V. F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjrner, A. P. Ershov, and N. D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531-549. North-Holland, 1988.

No context found.

--325. Turchin, V. F. 1988. The algorithm of generalization in the supercompiler. In Partial Evaluation and Mixed Computation, D. Bjrner, A. P. Ershov, and N. D. Jones, Eds. North-Holland, 531--

No context found.

V. Turchin. The algorithm of generalization in the supercompiler. In D. Bjrner, A. Ershov, and N. Jones, editors, Proc. of the IFIP TC2 Workshop on Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

No context found.

V.F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner et al., editors. Partial Evaluation and Mixed Computation. 341--353 (North-Holland, 1988).

No context found.

Valentin F. Turchin. The Algorithm of Generalization in the Supercompiler. In Partial Evaluation and Mixed Computation. Eds. A. P. Ershov, D. Bjrner & N. D. Jones North-Holland, 1988.

No context found.

V.F. Turchin. The algorithm of generalization in the supercompiler. In D. Bjørner, A.P. Ershov and N.D. Jones, editors, Partial Evaluation and Mixed Computation, pages 531--549. North-Holland, 1988.

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