M.H. Sørensen. Turchin's Supercompiler Revisited. An Operational Theory of Positive Information Propagation. Master's Thesis, DIKU report 94/7, 1994. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
A Positive Supercompiler - Sørensen, Glück, Jones (1993) (3 citations) (Correct)
....and we explain how hand rewriting the matcher using an idea due to Consel and Danvy (1989) solves the problem. 6.1 Pattern matching with the positive supercompiler Applying the positive supercompiler P to the term match AAB ss 0 returns the almost optimal program in Figure 11. 16 M.H. S rensen, R. Gluck, and N.D. Jones loopAAB ss0 loopAAB ss = case ss of [ F alse (s 0 : ss 0 ) if A = s 0 then loopAB ss 0 else loopAAB ss 0 loopAB ss = case ss of [ F alse (s 0 : ss 0 ) if A = s 0 then loopB ss 0 else if A = s 0 then loopAB ss 0 else loopAAB ss 0 loopB ss = case ss of [ F alse (s ....
Sørensen, M.H. 1994a. Turchin's supercompiler revisited. An operational theory of positive information propagation. Master's Thesis, DIKU-rapport 94/9, Department of Computer Science, University of Copenhagen.
An Algorithm of Generalization in Positive Supercompilation - Sørensen, Glück (1995) (1 citation) Self-citation (Srensen) (Correct)
....that is strictly stronger than partial evaluation and deforestation. It is capable of theorem proving and program inversion [20, 21, 22] and of program optimization beyond partial evaluation and deforestation [5] Recent renewed interest in supercompilation has lead to the positive supercompiler [6, 18, 19], a simplified version of Turchin s supercompiler for a functional language with trees as data structures. Supercompilation consists of driving and generalization, a technique to ensure termination of driving. A termination technique for the original supercompiler exists [24] The present paper ....
....A of the pattern to s. Building the partial process tree for, e.g. the term match [A; A; B] s and subsequently generating a new term and program gives a more effcient program with the same complexity as the specialized matchers produced by the well known Kuth Morris Pratt algorithm, as shown in [18]. This is a classical problem in program transformation; for references see [19] Neither partial evaluation nor deforestation can achieve this transformation, as explained in [19] Partial deduction achieves the same effect in the logic programming context [4, 17] as driving in the functional ....
[Article contains additional citation context not shown here]
M.H. Sørensen. Turchin's Supercompiler Revisited. An Operational Theory of Positive Information Propagation. Master's Thesis, DIKU report 94/7, 1994.
Partial Deduction and Driving are Equivalent - Glück, Sørensen (1994) (6 citations) Self-citation (Srensen) (Correct)
....the same as the I sequence. The operation of computing T trees is called driving. The description of Ttrees and T graphs here follows the description of process trees and graphs, respectively, as described in [Glu93] which in turn stem from graphs of states in Turchin s papers [Tur80a] See also [Sor94b]. We now show how one can derive a residual term t 0 and a residual program p 0 from a finite T graph. The basic idea is that every term e[r] in the T graph, where r is a f or g function call, gives rise to a new function definition, and the body, or bodies, of the new definition can be ....
....a version of the former into a version of the latter program: f(S x; r) g(x; r) f(S Z; Z) f(y; y) f(y; y) g(Z; Z) The two programs have the same computed answers, but some branches are pruned earlier in the SLD tree for the latter program. Related problems are explained in Section 9. 5 of [Sor94b] following [Tur80a] Example 14. Now let us instead compare the second of the functional programs from Example 5 with the same logic program. First note that the functional residual program has retained none of the functions from the original functional program from Example 1. This always holds. ....
[Article contains additional citation context not shown here]
M. H. Sørensen. Turchin's Supercompiler Revisited. An Operational Theory of Positive Information Propagation. Master's Thesis, Department of Computer Science, University of Copenhagen, 1994, Technical report 94/7.
Towards Unifying Partial Evaluation, Deforestation.. - Sørensen, Glück, Jones (1994) Self-citation (Srensen) (Correct)
....of being imprecise on this point. 6 As a matter of technicality, the patterns p g j and terms t g;c j in clause (5) must actually be chosen as renamings of the corresponding patterns and bodies for g, and in clause (8) the unifier must be chosen idempotent which is always possible, see [Sor94b]. This very important difference implies that in W we are propagating more information: the information that v has been instantiated. It will turn out that this accounts for the fact that W, but not S, is able to derive KMP style pattern matchers. It is easy to see that the algorithms have ....
....termination There are three issues of correctness for W: preservation of operational semantics, nondegradation of efficiency, and termination. First, the output of W should be semantically equivalent to the input. A proof of preservation of operational semantics, in a reasonable sense, is given in [Sor94b]. Second, the output of W should be at least as efficient as the input. Since rewriting to a nonlinear right hand side can cause function call duplication, this will not generally hold unless appropriate precautions are taken. The problem is well known in partial evaluation [Ses88, Bon90, Jon93] ....
[Article contains additional citation context not shown here]
M. H. Sørensen. Turchin's Supercompiler Revisited. An Operational Theory of Positive Information Propagation. Master's Thesis, Department of Computer Science, University of Copenhagen, 1994.
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