Two-Dimensional Bi-directional Object Layout

Proceedings of the 24th ACM SIGPLAN Conference on Object Oriented Programming Systems Languages and Applications

Morandat

Privat

2009

Object-oriented languages involve a threefold tradeoff between runtime efficiency, expressiveness (multiple inheritance), and modularity, i.e. open-world assumption (OWA). Runtime efficiency is conditioned by both the implementation technique and compilation scheme. The former specifies the data structures that support method invocation, attribute access and subtype testing. The latter consists of the production line of an executable from the source code. Many implementation techniques have been proposed and several compilation schemes can be considered from fully global compilation under the closed-world assumption (CWA) to separate compilation with dynamic loading under the OWA, with midway solutions. This article reviews a significant subset of possible combinations and presents a systematic, empirical comparison of their respective efficiencies with all other things being equal. The testbed consists of the PRM compiler that has been designed for this purpose. The considered techniques include C++ subobjects, coloring, perfect hashing, binary tree dispatch and caching. A variety of processors were considered. Qualitatively, these first results confirm the intuitive or theoretical abstract assessments of the tested approaches. As expected, efficiency increases as CWA strengthens. From a quantitative standpoint, the results are the first to precisely compare the efficiency of techniques that are closely associated with specific languages like C++ and EIFFEL. They also confirm that perfect hashing should be considered for implementing JAVA and .NET interfaces.

Section: Accessor Simulation (As)mentioning

confidence: 99%

Empirical assessment of object-oriented implementations with multiple inheritance and static typing

Proceedings of the 24th ACM SIGPLAN Conference on Object Oriented Programming Systems Languages and Applications

Morandat

Privat

2009

“…Coloring is a versatile implementation technique whose different variants have been discovered and rediscovered by several people [Dixon et al, 1989 ;Pugh et Weddell, 1990 ;Ducournau, 1991 ;Zibin et Gil, 2003]. The main goal of this paper was to generalize these different works and highlight their unity-i.e.…”

Section: Discussionmentioning

confidence: 98%

“…Bidirectional coloring involves both positive and negative colors, contrary to unidirectional coloring, which involves only positive colors. Generalization to n-directional or n-dimensional coloring was further proposed by [Pugh et Weddell, 1993] and rediscovered by [Zibin et Gil, 2003]- [Gil et al, 2008] is a synthesis. Besides coloring, the bidirectionality of method tables or object layout has been widely reused [Myers, 1995 ;Krall et Grafl, 1997 ;Gil et Sweeney, 1999 ;Gagnon et Hendren, 2001].…”

Section: Bidirectional and N-dimensional Coloringmentioning

confidence: 99%

“…However, a class can simulate accessors by replacing the method address in the method table with the attribute's offset. This approach is called field dispatching by [Zibin et Gil, 2003]. Another improvement is to group attributes introduced by the same class together in the method table.…”

Section: Accessors and Accessor Simulationmentioning

confidence: 99%

“…These subobjects might be explicitly linked with n−1 pointers in the first subobject, but it would contradict the intention of saving on space-this would generate the same drawbacks as VBPTRs (Section 2.2.2). Therefore, the alternative proposed by [Zibin et Gil, 2003] would object table object table object method table object Figure 10: n-dimensional object layout, with an n-pointer header in the object layout (left) or with indirections through offsets in the method table (right).…”

Section: Application Of N-dimensional Coloringmentioning

confidence: 99%

See 2 more Smart Citations

Coloring, a versatile technique for implementing object‐oriented languages

2010

Softw Pract Exp

Late binding and subtyping create run-time overhead for object-oriented languages. Dynamic typing and multiple inheritance create even more overhead. Static typing and single inheritance lead to two major invariants-of reference and position-that make the implementation as efficient as possible. Coloring is a technique that preserves these invariants for dynamic typing or multiple inheritance at minimal spatial cost.Coloring has been introduced and applied, more or less independently, to method invocation-under the name of selector coloring-to subtype tests under the name of pack encoding and to attribute access. This paper reviews a number of uses of coloring for optimizing object-oriented programming and generalizes them. It specifies several variations of coloring, such as bidirectional coloring and n-dimensional coloring. Coloring is NP-hard, so compilers that use it depend on heuristics. The paper describes some experimental results which indicate that bidirectional coloring gives the best results.

Perfect class hashing and numbering for object‐oriented implementation

Morandat

2010

Softw Pract Exp

Late binding and subtyping create run‐time overhead for object‐oriented languages, especially in the context of both multiple inheritance and dynamic loading, for instance for JAVA interfaces. In a previous article, we proposed a novel approach based on perfect hashing and truly constant‐time hashtables for implementing subtype testing and method invocation in a dynamic loading setting. In this first study, we based our efficiency assessment on Driesen's abstract computational model for the time aspect, and on large‐scale benchmarks for the space aspect. The conclusions were that the technique was promising but required further research in order to assess its scalability. This article presents some new results on perfect class hashing that enhance its interest. We propose and test both new hashing functions and an inverse problem that amounts to selecting the best class identifiers in order to minimize the overall hashtable size. This optimizing approach is proven to be optimal for single‐inheritance hierarchies. Experiments within an extended testbed with random class loading and under cautious assumptions about what should be a sensible class‐loading order show that perfect class hashing scales up gracefully, especially on JAVA‐like multiple‐subtyping hierarchies. Furthermore, perfect class hashing is implemented in the PRM compiler testbed, and compared here with the coloring technique, which amounts to maintaining the single‐inheritance implementation in multiple inheritance. The overall conclusion is that the approach is efficient from both time and space standpoints with the bit‐wise and hashing function. In contrast, the poor time efficiency of modulus hashing function on most processors is confirmed. Copyright © 2010 John Wiley & Sons, Ltd.