Lightweight closure conversion

Steckler, Paul; Wand, Mitchell

doi:10.1145/239912.239915

Cited by 45 publications

(11 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(1) uncurrying [19], which can increase the number of closed functions; (2) closure representations that exclude free variables from an environment if their values are available at all call sites [36]; and (3) register allocation and calling conventions informed by flow information. There are numerous other closure representation tricks (e.g., those discussed in [24,2,34]) to investigate in the context of our framework.…”

Section: Discussionmentioning

confidence: 99%

Functioning without closure

et al. 2001

View full text Add to dashboard Cite

The CIL compiler for core Standard ML compiles whole ML programs using a novel typed intermediate language that supports the generation of type-safe customized data representations. In this paper, we present empirical data comparing the relative efficacy of several different flow-based customization strategies for function representations. We develop a cost model to interpret dynamic counts of operations required for each strategy. In this cost model, customizing the representation of closed functions gives a 12-17% improvement on average over uniform closure representations, depending on the layout of the closure. We also present data on the relative effectiveness of various strategies for reducing representation pollution, i.e., situations where flow constraints require the representation of a value to be less efficient than it would be in ideal circumstances. For the benchmarks tested and the types of representation pollution detected by our compiler, the pollution removal strategies we consider often cost more in overhead than they gain via enabled customizations. Notable exceptions are selective defunctionalization, a function representation strategy that often achieves significant customization benefits via aggressive pollution removal, and a simple form of flow-directed inlining, in which pollution removal allows multiple functions to be inlined at the same call site.

show abstract

Section: Discussionmentioning

confidence: 99%

Functioning without closure

et al. 2001

View full text Add to dashboard Cite

show abstract

“…Steckler and Wand [17] describe a closure-conversion algorithm that creates "light-weight closures" that do not contain free variables that are available at the call site. This is a limited form of lambda lifting and, as with full lambda lifting, can sometimes do harm relative to the straight flat-closure model.…”

Section: Related Workmentioning

confidence: 99%

“…While a few of the optimizations have long been performed by our compiler, descriptions of them have never been published. Various closure optimizations have been described by others [2,6,8,[11][12][13][14]17], but most of the optimizations described here have not been described previously in the literature, and many are likely novel.…”

Section: Introductionmentioning

confidence: 99%

Optimizing closures in O(0) time

Keep¹,

Hearn

Dybvig

2012

Proceedings of the 2012 Annual Workshop on Scheme and Functional Programming

View full text Add to dashboard Cite

The flat-closure model for the representation of first-class procedures is simple, safe-for-space, and efficient, allowing the values or locations of free variables to be accessed with a single memory indirect. It is a straightforward model for programmers to understand, allowing programmers to predict the worst-case behavior of their programs. This paper presents a set of optimizations that improve upon the flat-closure model along with an algorithm that implements them, and it shows that the optimizations together eliminate over 50% of run-time closure-creation and free-variable access overhead in practice, with insignificant compile-time overhead. The optimizations never add overhead and remain safe-for-space, thus preserving the benefits of the flat-closure model.

show abstract

“…For any first-order imperative program, such information is available via a flowchart constructed from the program text. Given this information, the call-graph of a program can be constructed and a suite of compiler optimisations, for example, closure conversion (Steckler and Wand 1997;Dimock et al 2001), useless variable elimination (Wand and Siveroni 1999), constant propagation, induction variable elimination (Shivers 1991), and so on, can be enabled. For higher-order programs, however, a control-flow graph is often not evident from the program text.…”

Section: Introductionmentioning

confidence: 99%

Modular control-flow analysis with rank 2 intersection types

Banerjee

Jensen

2003

Math. Struct. in Comp. Science

View full text Add to dashboard Cite

We show how the principal typing property of the rank 2 intersection type system enables the specification of a modular and polyvariant control-flow analysis. † Partially supported by postdoctoral fellowships at technique uses a system of constraints to specify control-flow analysis so that flow information is obtained as the minimal solution of the constraint system (Palsberg 1995). Both techniques require whole-program analysis. For control-flow analysis of program fragments containing free variables, the above techniques usually assume an environment that associates a property with each free variable or assume that only trivial properties hold for the free variables. Neither assumption is satisfactory: the former requires reanalysis whenever the environment changes; the latter leads to poor analysis results and the rejection of program fragments that require functions as inputs.We describe an algorithm for modular and polyvariant control-flow analysis of simply typed program fragments. The result of analysing a program fragment P is a pair containing a property and an environment: the pair describes a relation between the properties of the free variables of P and the property of the entire program fragment P . The environment provides a summary of the minimum set of constraints that must be satisfied by any other program fragment that may link to P . Our algorithm computes a 'principal solution': the environment and property of any other program fragment that links to P can be obtained as an instance of the environment and property computed for P (see Theorem 6.4 for the precise statement). Thus, no re-analysis of P is required.Recently, there has been much interest in using annotated type systems for program analysis. The intuition is that types and expressions can be annotated with the properties of interest, for exampleso on, so that if an expression e has the annotated type τ, then evaluation of the expression exhibits the properties described by the annotation of τ. Static analysis of the expression e is then synonymous with the calculation, via an annotated-type inference algorithm, of its property annotations. For control-flow analysis, we annotate every function in an expression with a label, and associate a set of function labels ϕ with every function type τ. The intuition is that if e evaluates to a closure, then the text of the closure is a function whose label is in ϕ. The static determination of the set of function labels that e can possibly evaluate to is thus synonymous with the calculation, via an inference algorithm, of its flow annotations.An advantage of the type-based approach is that it provides a method for performing compositional and modular program analysis. 'Compositional' means that the analysis of an expression is derived through the composition of the analyses of its proper subparts. We say that an analysis is 'modular' if it can analyse program fragments containing free variables in isolation, and if the linking of fragments does not require their re-analysis. A modular program an...

show abstract

Lightweight closure conversion

Cited by 45 publications

References 26 publications

Functioning without closure

Functioning without closure

Optimizing closures in O(0) time

Modular control-flow analysis with rank 2 intersection types

Contact Info

Product

Resources

About