S. Finne and G. Burn. Assessing the Evaluation Transformer Model of Reduction on the Spineless G-machine. In FPCA'93 [FPC93], pages 331340.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....be applied to sequential as well as to parallel implementations of lazy functional languages. The extra cost for runtime evaluator information seems to be useful only in parallel implementations. Some work to implement lazy functional languages using evaluation transformers has already been done [1, 3, 5, 6, 8], but all these approaches mainly deal with evaluators for simple structured types like lists and trees. The problems which arise when using evaluators for general structured types are not addressed. This article defines a universally valid evaluator space containing evaluators for general ....

Sigbjrn Finne, Geoffrey L. Burn; Assessing the Evaluation Transformer Model of Reduction on the Spineless G-machine; In: Functional Programming & Computer Architecture, ACM, June 1993

....lists will be evaluated completely, i.e. evaluation to normal form. The evaluation transformer analysis 1 seems to produce many more computable expressions than simple strictness analysis. Some work to implement lazy functional languages using evaluation transformers has already been done [1, 6, 9, 11, 13]. But all these approaches mainly deal with evaluators for simple structured types like lists and trees. The problems arising when using evaluators for general structured types are not addressed. One reason for this is the absence of practicable analysis methods for this general case. 18] deals ....

Sigbjrn Finne, Geoffrey L. Burn; Assessing the Evaluation Transformer Model of Reduction on the Spineless G-machine; In: Functional Programming & Computer Architecture, ACM, June 1993

....lists will be evaluated completely, i.e. evaluation to normal form. The evaluation transformer analysis 1 seems to produce many more computable expressions than simple strictness analysis. Some work to implement lazy functional languages using evaluation transformers has already been done [1, 6, 8, 10, 12]. But all these approaches mainly deal with evaluators for simple structured types like lists and trees. The problems arising when using evaluators for general structured types are not addressed. One reason for this is the absence of practicable analysis methods for this general case. 17] deals ....

Sigbjrn Finne, Geoffrey L. Burn; Assessing the Evaluation Transformer Model of Reduction on the Spineless G-machine; In: Functional Programming & Computer Architecture, ACM, June 93

....the degree of evaluation that is required of the argument given a degree of evaluation required of the application. While there has been some research on the exploitation of strictness information for optimization e.g. PJS94, HB93, PJL91, PJP] research on context information and its use e.g. [LB96, FB94, Sew92, Far92] has not received much attention. Many approaches to strictness analysis define a finite set of abstract values for any particular type beforehand for use in the analysis, e.g. Wadler s 4 point This work was supported by the Deutsche Forschungsgemeinschaft under Az Schm 986 4 1 1 domain [Bur91, ....

....similar to lazy evaluation and does not compute information, which is not explicitly requested or needed to obtain the requested information. To use the results of our calculus for optimization the compiler could allow the user to specify optimization heuristics, i.e. evaluation transformers [FB94, Bur91], or tell the compiler how to use an optimized data representation as in [Hal94] In the terminology of [CC77] our approach would be called abstract interpretation, yet subsequent work by [Myc80, BHA85, Bur87, Bur91] and others used the term abstract interpretation almost exclusively for abstract ....

Sigbjrn Finne and Geoffrey Burn. Assessing the evaluation transformer model of reduction on the Spineless G-machine. Technical report, Imperial College of Science, Technology and Medicine, Department of Computing, 1994.

....only. As will be seen in section 4 the context analysis problem is easily reduced to the strictness analysis problem. Today most compilers do not use context information for code generation and the question of how to gain efficiency by context information is at least controversial, see e.g. FB94] Nevertheless, we made some tests for context analysis, see section 5. For a detailed discussion of the implementational issues and the complete operational semantics of the G # machine the reader is referred to [Sch95] 3.1 Open Redexes Reduction path analysis makes it necessary to ....

Sigbjrn Finne and Geoffrey Burn. Assessing the evaluation transformer model of reduction on the Spineless G-machine. Technical report, Imperial College of Science, Technology and Medicine, Department of Computing, 1994.

....strictness properties (Burn, 1987) Each evaluation transformer reduces its argument to the extent that is allowed by the available strictness information. The appropriate transformer is selected at compile time, giving efficient execution at the cost of some increase in code size (Burn, 1991; Finne and Burn, 1993). If there are only a small number of possible transformers (as for lists using the standard 4 point strictness domain see Table 1) repeated work can be avoided by recording the extent to which a data structure has already been evaluated, and then using a specialised transformer on the ....

Finne, S.O., and Burn, G.L., "Assessing the Evaluation Transformer Model of Reduction on the Spineless G-Machine", Proc. FPCA '93, Copenhagen, (1993), pp. 331--340.

....strictness properties (Burn, 1987) Each evaluation transformer reduces its argument to the extent that is allowed by the available strictness information. The appropriate transformer is selected at compile time, giving efficient execution at the cost of some increase in code size (Burn, 1991; Finne and Burn, 1993). If there are only a small number of possible transformers (as for lists using the standard 4 point strictness domain see Table 1) repeated work can be avoided by recording the extent to which a data structure has already been evaluated, and then using a specialised transformer on the ....

Finne, S.O., and Burn, G.L., "Assessing the Evaluation Transformer Model of Reduction on the Spineless G-Machine", Proc. FPCA '93, Copenhagen, (1993), pp. 331--340.

....but two aspects of the problem have received little attention until recently: 1. The integration of the results of the analysis into a real compiler. 2. The efficiency of the algorithm implementing the analysis. The first issue has been tackled recently both from an experimental point of view [11, 16] and from a theoretical point of view [5, 8] We are concerned with the second issue in this paper. The abstract interpretation and the projections approaches have led to the construction of analyses based on rich domains which make them intractable even for some simple examples. Techniques ....

S. Finne and G. Burn, Assessing the evaluation transformer model of reduction on the spineless G-machine, in Proceedings of the 6th ACM Conference on Functional Programming Languages and Computer Architecture, ACM Press, 1993, pp. 331-341.

....for a spine strict context will compute the whole of the result before returning any of it, which might have a terrible effect on the peak space requirement of the program. There is very little published work which attempts to measure the effects of exploiting strictness in recursive data types. Finne Burn [1993] use evaluation transformers to evaluate lists more strictly (in the same spirit as the work described here) but for sequential implementations they find few performance benefits, and occasional large costs. In contrast, Hall [1993] uses strictness information over list types to guide a new ....

S Finne & G Burn [June 1993], "Assessing the evaluation transformer model of reduction on the spineless G-machine," in Proc Functional Programming Languages and Computer Architecture, Copenhagen, ACM, 331--340.

....i.e. they can always obtain almost linear speedup. However, there is an important weakness in this conclusion: none of their program really require lazy evaluation, that is, none exploit the major feature of lazy evaluation, namely, the ability to manipulate potentially infinite objects. In [16], an assessment of the evaluation transformer model of reduction [11] on the performance of lazy programs is presented. Although various parameters are examined, e.g. heap allocation, number of memory accesses, etc. parallelism per se is not examined. Note that one of their interesting ....

S. Finne and G. Burn. Assessing the evaluation transformer model of reduction on the spineless G-machine. In FPCA '93, pages 331--340. ACM Press, 1993.

....for projection analysis. We will not discuss the analysis here, but refer the reader to a technical report [30] Instead we will show the results of the analysis. The code that is generated by the compiler for functions to be called in the various context styles, implements compile time choice [31, 32] of versions. This means that for each possible context (1,2,3 etc. the compiler creates a version of those functions that are capable of delivering the right style of list. For Append there is version that produces a head normal form but also a version that delivers a spine. The C code generated ....

S. O. Finne and G. L. Burn. Assessing the evaluation transformer model of reduction on the spineless G-machine. In 6th Functional programming languages and computer architecture, pages 331--340, Copenhagen, Denmark, Jun 1993. ACM, New York.

....require of an evaluation context [Bur91a] Definition 3.1 (Scott closed set) A set S is Scott closed of a domain D if 1. it is down closed, that is, if 8d 2 D such that 9s 2 S such that d v s, then d 2 S; and 1 Results from some initial experiments investigating this question can be found in [FB93]. T V i Q j [ x] x if x 62 V = x ij if x 2 V T V i Q j [ k oe ] N [ k oe ] if k oe 6= if T V i Q j [ if E 1 E 2 E 3 ] c:T V 0 f S bool g [ E 1 ] if c (T V i Q j [ E 2 ] c) T V i Q j [ E 3 ] c) T V i Q j [ E 1 E 2 ] c:T V 0 P [ E 2 ] v:T V (i 1) Q j [ E 1 ] f:f c ....

S.O. Finne and G.L. Burn. Assessing the evaluation transformer model of reduction on the spineless g-machine. In Proceedings of the Conference on Functional Programming and Computer Architecture, pages 331--340, ACM, Copenhagen, 9--11 June 1993.

No context found.

S. Finne and G. Burn. Assessing the Evaluation Transformer Model of Reduction on the Spineless G-machine. In FPCA'93 [FPC93], pages 331340.

No context found.

S. O. Finne and G. L. Burn, `Assessing the evaluation transformer model of reduction on the spineless G-machine', in 6th Functional Programming Languages and Computer Architecture, Copenhagen, Denmark, June 1993. ACM, pp. 331--340.

No context found.

S. Finne and G. Burn, Assessing the evaluation transformer model of reduction on the spineless G-machine, in Proceedings of the 6th ACM Conference on Functional Programming Languages and Computer Architecture, ACM Press, 1993, pp. 331-341.

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