H. Seidl and M.H. Srensen. Constraints to Stop Higher-Order Deforestation. In N.D. Jones, editor, ACM Symposium on Principles of Programming Languages, pages 400-- 413, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....data structure for use by f, fusion would attempt to merge the composition into a specialised function p(x) with the same semantics as f(g(x) but without the need for an intermediate data structure. In recent years, many attempts have been put forward to automate such fusion calculations [SF93,GLPJ93,SS97,HIT96]. Most current attempts are restricted to using equational de nitions during transformation. For example, the deforestation algorithm [Wad88] relies on only de ne, unfold and fold rules [BD77] base on equational de nitions. Unfortunately, this approach is inadequate since we often need laws ....

H. Seidl and M.H. Srensen. Constraints to stop higher-order deforestation. In 24th ACM Symposium on Principles of Programming Languages, Paris, France, January 1997. ACM Press.

.... constraints to represent the two level expressions [Hen91] In deforestation, an analysis to detect unsafe expressions approximates the set of terms encountered during the transformation by set constraints [Sei96] This technique was refined to ensure the termination of higher order deforestation [SS97] Constraint solving has also been applied to the problem of propagating information during the transformation. For example, in supercompilation, negative information is propagated by restrictions, which can be represented by Herbrand (term) constraints [Tur86, GK93] The ability to propagate ....

H. Seidl and M.H. Srensen. Constraints to Stop Higher-Order Deforestation. In N.D. Jones, editor, ACM Symposium on Principles of Programming Languages, pages 400-- 413, 1997.

....data structure for use by f; fusion would attempt to merge the composition into a specialised function p(x) with the same semantics as f(g(x) but without the need for an intermediate data structure. In recent years, many attempts have been put forward to automate such fusion calculations [SF93, GLPJ93, SS97, HIT96]. Most current attempts to automate fusion are restricted to the use of equational definitions of functions to perform transformation. For example, the deforestation algorithm [Wad88] relies on only define, unfold and fold rules [BD77] in its transformation which can be carried out using ....

H. Seidl and M.H. Srensen. Constraints to stop higher-order deforestation. In 24th ACM Symposium on Principles of Programming Languages, Paris, France, January 1997. ACM Press.

....suitable as a compiler optimisation. 1 Introduction Source to source transformation can achieve global optimisation through specialisation for recursive functions. Two well known techniques are partial evaluation [9] and deforestation [20] Both techniques have been extensively investigated [18, 8] to discover automatic algorithms and supporting analyses that can ensure correct and terminating program optimisations. Tupling is a lesser known but equally powerful transformation technique. The basic technique works by grouping calls with common arguments together, so that their multiple ....

....techniques. Though the problem of determining termination is in general undecidable, a variety of analyses can be applied to give meaningful results. In the case of deforestation, the proposals range from simple pure treeless syntactic form [20] to a sophisticated constraint based analysis [18] 5 to stop the transformation. Likewise, earlier tupling work [2, 7] were based simply on restricted functions. In [2] the transformable functions can only have a single recursion parameter each, while accumulative parameters are forbidden. Similarly, in the calculational approach of [7] the ....

H. Seidl and M.H. Srensen. Constraints to stop higher-order deforestation. In 24th ACM Symposium on Principles of Programming Languages, Paris, France, January 1997. ACM Press.

....was shown to subsume Chin s earlier approach. ffl Marlow [Mar95] and Hamilton [Ham93] separately extended deforestation to higher order programs. An extended form of treelessness, together with finite typing, is used to guarantee that higher order deforestation terminates. ffl Seidl and Srensen [SS96] recently proposed a new constraint based analysis to stop higher order deforestation. An extra unsafe feature, called accumulating spine, is identified to avoid nontermination caused by certain higher order programs. The alternative on line approach imposes periodic checks at various program ....

....plus on line generalizations. Our approach is considered off line, but the simplicity of our analyses is compensated by the aggressiveness of the transformation strategy. The result of final specialization appears to be comparable to the more sophisticated semantics based approaches (e.g. [Sr94, SS96]) with the ability to handle essentially the same range of programs. However, the semantics based analyses of Srensen and Seidl [Sr94, Sei96] which mirror deforestation by analyzing the control flow starting from some top most expression, suffers from two weaknesses. First, this top down ....

[Article contains additional citation context not shown here]

H. Seidl and M. H. Srensen. Constraints to stop higher-order deforestation. Submitted, 1996.

....for efficiency. To get the best of both worlds, deforestation transformations were introduced. These transformations fuse two pieces of a program into another one, where intermediate data structure constructions have been eliminated. The first approach dealing with such transformation is Wadler s [32, 3, 29]. It is based on the fold and unfold transformation [2] There is another interesting approach based on algebraic notions often called deforestation in calculational form [14, 30, 21, 31, 15] The idea of this approach is to capture both the function and the data type patterns of recursion ....

H. Seidl and M.H. Sørensen. Constraints to stop higher-order deforestation. In 24th ACM Symp. on Principles of Programming Languages, pages 400--413, 1997.

.... of imperative programs and applied to interprocedural linear constant propagation [17] A variant of control flow analysis ( set based analysis ) has been proposed by Heintze for the analysis of ML [13] Another variant, even more in the spirit of the methods used here, has been applied in [26] to a higher order functional language with call by need semantics to obtain a termination analysis for deforestation. In section 12 we extend applicability of algorithm WR Delta further by introducing generalized computation trees. This generalization takes into account independence of ....

.... Control flow analysis (cfa for short) is an analysis for higher order functional languages possibly with recursive data types [13] Cfa on program p tries to compute for every expression t occurring in p a superset of expressions into which t may evolve during program execution, see, e.g. [23, 24, 21, 26]. Let A denote the set of subexpressions of p and D = 2 A . Then cfa for a lazy language as in [26] can be formalized through a constraint system S over domain D with set V of in variables y t ; t 2 A, where righthand sides of constraints consist of expressions e of one of the following forms: ....

[Article contains additional citation context not shown here]

H. Seidl and M.H. Srensen. Constraints to Stop Higher-Order Deforestation. 24th POPL, 400--413, 1997.

.... of imperative programs and applied to interprocedural linear constant propagation [16] A variant of control flow analysis ( set based analysis ) has been proposed by Heintze for the analysis of ML [12] Another variant, even more in the spirit of the methods used here, has been applied in [25] to a higher order functional language with call by need semantics to obtain a termination analysis for deforestation. In Section 12 we extend the applicability of algorithm WR Delta further by introducing generalized computation trees. This generalization takes into account independence of ....

.... Control flow analysis (cfa for short) is an analysis for higher order functional languages possibly with recursive data types [12, 13] Cfa on program p tries to compute for every expression t occurring in p a superset of expressions into which t may evolve during program execution, see, e.g. [22, 23, 20, 25]. Let A denote the set of subexpressions of p and D = 2 A . Then cfa for a lazy language as in [25] can be formalized through a constraint system S over domain D with set V of in variables y t ; t 2 A, where right hand sides of constraints consist of expressions e of one of the following forms: ....

[Article contains additional citation context not shown here]

Helmut Seidl and Morten Heine Srensen. Constraints to Stop Higher-Order Deforestation. In ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), pages 400--413, 1997.

.... of imperative programs and applied to interprocedural linear constant propagation [18] A variant of control flow analysis ( set based analysis ) has been proposed by Heintze for the analysis of ML [14] Another variant, even more in the spirit of the methods used here, has been applied in [26] to a higher order functional language with call by need semantics to obtain a termination analysis for deforestation. In section 12 we extend applicability of algorithm WR Delta further by introducing generalized decision trees. This generalization takes into account independence of ....

.... analysis (cfa for short) is an analysis for higher order functional languages possibly with recursive data types [14, 15] Cfa on program p tries to compute for every expression t occurring in p a superset of expressions into which t may evolve during program execution [22] It is used, e.g. in [26] as the basis of a termination analysis for deforestation (see also [23, 24] Let A denote the set of subexpressions of p, and V denote the set of variables y t , t 2 A. Furthermore, let D denote the complete lattice of all subsets of A. Then cfa for p can be formalized through a constraint ....

Helmut Seidl and Morten Heine Sørensen. Constraints to Stop Higher-Order Deforestation. In ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), pages 400--413, 1997.

....[11] shows that some shortcomings of the treeless syntax can be avoided by ad hoc extensions of deforestation. The necessity of such extensions stems from the fact that the blazing scheme is syntactic; it does not take into account what happens during deforestation. In contrast, Seidl and S rensen [52] give a single technique coping with many of these problems by generalizing their constraint based control flow analysis [59,50] to work for higher order deforestation. The technique is partly conservative over the ones based on the higher order treeless syntax: for a certain large class of ....

H. Seidl and M.H. Sørensen. Constraints to stop higher-order deforestation. In Conference Record of the Annual ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages, pages 400--413. ACM Press, 1997.

No context found.

H. Seidl and M.H. Srensen. Constraints to Stop Higher-Order Deforestation. In N.D. Jones, editor, ACM Symposium on Principles of Programming Languages, pages 400-- 413, 1997.

No context found.

H. Seidl, M.H. Srensen. Constraints to Stop Higher-Order Deforestation In 24th ACM Symp. on Principles of Programming Languages, pages 400--413, 1997.

No context found.

H. Seidl and M. H. S#rensen. Constraints to Stop HigherOrder Deforestation. In POPL'97 [POP97], pages 400413.

No context found.

H. Seidl and M.H. Srensen. Constraints to stop higher-order deforestation. In POPL'97, Paris, pages 400--413. ACM Press, 1997.

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