The last years have witnessed a flurry of new results in the area of partial evaluation. These tutorial notes survey the field and present a critical assessment of the state of the art. 1 Introduction Partial evaluation is a source-to-source program transformation technique for specializing programs with respect to parts of their input. In essence, partial evaluation removes layers of interpretation. In the most general sense, an interpreter can be defined as a program whose control flow is determined by its input data. As Abelson points out, [43, Foreword], even programs that are not themselves interpreters have important interpreter-like pieces. These pieces contain both compile-time and run-time constructs. Partial evaluation identifies and eliminates the compile-time constructs. 1.1 A complete example We consider a function producing formatted text. Such functions exist in most programming languages (e.g., format in Lisp and printf in C). Figure 1 displays a formatting functio...

Object-oriented programming languages confer many benefits, including abstraction, which lets the programmer hide the details of an object’s implementation from the object’s clients. Unfortunately, crossing abstraction boundaries often incurs a substantial run-time overhead in the form of frequent procedure calls. Thus, pervasive use of abstraction, while desirable from a design standpoint, may be impractical when it leads to inefficient programs. Aggressive compiler optimizations can reduce the overhead of abstraction. However, the long compilation times introduced by optimizing compilers delay the programming environment‘s responses to changes in the program. Furthermore, optimization also conflicts with source-level debugging. Thus, programmers are caught on the horns of two dilemmas: they have to choose between abstraction and efficiency, and between responsive programming environments and efficiency. This dissertation shows how to reconcile these seemingly contradictory goals by performing optimizations lazily. Four new techniques work together to achieve high performance and high responsiveness: • Type feedback achieves high performance by allowing the compiler to inline message sends based on information extracted from the runtime system. On average, programs run 1.5 times faster than the previous SELF system; compared to a commercial Smalltalk implementation, two medium-sized benchmarks run about three times faster. This level of performance is obtained with a compiler that is both simpler and faster than previous SELF compilers. • Adaptive optimization achieves high responsiveness without sacrificing performance by using a fast nonoptimizing compiler to generate initial code while automatically recompiling heavily used parts of the program with an optimizing compiler. On a previous-generation workstation like the SPARCstation-2, fewer than 200 pauses exceeded 200 ms during a 50-minute interaction, and 21 pauses exceeded one second. On a currentgeneration workstation, only 13 pauses exceed 400 ms. • Dynamic deoptimization shields the programmer from the complexity of debugging optimized code by transparently recreating non-optimized code as needed. No matter whether a program is optimized or not, it can always be stopped, inspected, and single-stepped. Compared to previous approaches, deoptimization allows more debugging while placing fewer restrictions on the optimizations that can be performed. • Polymorphic inline caching generates type-case sequences on-the-fly to speed up messages sent from the same call site to several different types of object. More significantly, they collect concrete type information for the optimizing compiler. With better performance yet good interactive behavior, these techniques make exploratory programming possible both for pure object-oriented languages and for application domains requiring higher ultimate performance, reconciling exploratory programming, ubiquitous abstraction, and high performance.

Partial evaluators can be separated into two classes: offline specializers, which make all of their reduce/residualize decisions before specialization, and online specializers, which make such decisions during specialization. The choice of which method to use is driven by a tradeoff between the efficiency of the specializer and the quality of the residual programs that it produces. Existing research describes some of the inefficiencies of online specializers, and how these are avoided using offline methods, but fails to address the price paid in specialization quality. This paper motivates research in online specialization by describing two fundamental limitations of the offline approach, and explains why the online approach does not encounter the same difficulties.

A flow-directed inlining strategy uses information derived from control-flow analysis to specialize and inline procedures for functional and object-oriented languages. Since it uses control-flow analysis to identify candidate call sites, flowdirected inlining can inline procedures whose relationships to their call sites are not apparent. For instance, procedures defined in other modules, passed as arguments, returned as values, or extracted from data structures can all be inlined. Flow-directed inlining specializes procedures for particular call sites, and can selectively inline a particular procedure at some call sites but not at others. Finally, flow-directed inlining encourages modular implementations: control-flow analysis, inlining, and post-inlining optimizations are all orthogonal components. Results from a prototype implementation indicate that this strategy effectively reduces procedure call overhead and leads to significant reduction in execution time. 1 Introduction Functio...

This paper present a semantan(7W1W2 compilat2( framework for an aspect716C7 tt programming language based on it operat'W(/ semantma model. Usingpart(1 evaluatlua tl framework can explain several issues incompilat]2 processes, including howt o find places in programtog t insert aspect code and howt o remove unnecessaryrun-tes checks.It also illustW(/2 optstW(/2' of calling-contal sensit - e pointn(' (cflow), implementp in real compilers. Keywords Aspect SandBox, dynamic join point model,part12 evaluatu -( FutC ura project2C1 compile-t2] weaving, contn(1 sensit] e point(7 designat21 cflow) 1.

This paper gives a gentle introduction to Turchin's supercompilation and its applications in metacomputation with an emphasis on recent developments. First, a complete supercompiler, including positive driving and generalization, is defined for a functional language and illustrated with examples. Then a taxonomy of related transformers is given and compared to the supercompiler. Finally, we put supercompilation into the larger perspective of metacomputation and consider three metacomputation tasks: specialization, composition, and inversion.

We investigate the application of partial evaluation to numerically oriented computation in scientific and engineering applications. We present our results using the Fast Fourier Transformation, the N-body attraction problem, and the cubic splines interpolation as examples. All programs are written in Fortran 77, specialized using our Fortran partial evaluator, and compiled into executable machine code using a commercial Fortran compiler. The results demonstrate that existing partial evaluation technology is strong enough to improve the efficiency of a large class of numerical programs. However, using partial evaluation as a development tool in the ‘real world’ still remains a challenging problem.

Program specialization can divide a computation into several computation stages. This paper investigates the theoretical limitations and practical problems of standard specialization tools, presents multi-level specialization, and demonstrates that, in combination with the cogen approach, it is far more practical than previously supposed. The program generator which we designed and implemented for a higher-order functional language converts programs into very compact multi-level generating extensions that guarantee fast successive specialization. Experimental results show a remarkable reduction of generation time and generator size compared to previous attempts of multi-level specialization by self-application. Our approach to multi-level specialization seems well-suited for applications where generation time and program size are critical.

MULTIPLE SPECIALIZATION AND ITS APPLICATION TO PROGRAM PARALLELIZATION GERMAN PUEBLA AND MANUEL HERMENEGILDO . Program specialization optimizes programs for known values of the input. It is often the case that the set of possible input values is unknown, or this set is innite. However, a form of specialization can still be performed in such cases by means of abstract interpretation, specialization then being with respect to abstract values (substitutions), rather than concrete ones. We study the multiple specialization of logic programs based on abstract interpretation. This involves in principle, and based on information from global analysis, generating several versions of a program predicate for dierent uses of such predicate, optimizing these versions, and, nally, producing a new, \multiply specialized" program. While multiple specialization has received theoretical attention, little previous evidence exists on its practicality. In this paper we report on the incor...

We study the multiple specialization of logic programs based on abstract interpretation. This involves in general generating several versions of a program predicate for different uses of such predicate, making use of information obtained from global analysis performed by an abstract interpreter, and finally producing a new, "multiply specialized" program. While the topic of multiple specialization of logic programs has received considerable theoretical attention, it has never been actually incorporated in a compiler and its effects quantified. We perform such a study in the context of a parallelizing compiler and show that it is indeed a relevant technique in practice. Also, we propose an implementation technique which has the same power as the strongest of the previously proposed techniques but requires little or no modification of an existing abstract interpreter. Keywords: Multiple Program Specialization, Abstract Interpretation, Logic Programming, Compile-time Analysis, Optimizati...