FromPushEntertoEvalApplyby.PDFVIP

From Push/Enter to Eval/Apply by Program Transformation Maciej Pir′og Jeremy Gibbons Department of Computer Science University of Oxford maciej.pirog@cs.ox.ac.uk jeremy.gibbons@cs.ox.ak Push/enter and eval/apply are two calling conventions used in implementations of functional lan- guages. In this paper, we explore the following observation: when considering functions with mul- tiple arguments, the stack under the push/enter and eval/apply conventions behaves similarly to two particular implementations of the list datatype: the regular cons-list and a form of lists with lazy concatenation respectively. Along the lines of Danvy et al.’s functional correspondence between def- initional interpreters and abstract machines, we use this observation to transform an abstract machine that implements push/enter into an abstract machine that implements eval/apply. We show that our method is ?exible enough to transform the push/enter Spineless Tagless G-machine (which is the semantic core of the GHC Haskell compiler) into its eval/apply variant. 1 Introduction There are two standard calling conventions used to ef?ciently compile curried multi-argument functions in higher-order languages: push/enter (PE) and eval/apply (EA). With the PE convention, the caller pushes the arguments on the stack, and jumps to the function body. It is the responsibility of the function to ?nd its arguments, when they are needed, on the stack. With the EA convention, the caller ?rst evaluates the function to a normal form, from which it can read the number and kinds of arguments the function expects, and then it calls the function body with the right arguments. The difference between the two calling conventions is thoroughly discussed by Marlow and Peyton Jones [12]. The terms PE and EA are also used in the literature to descri