de.mpg.escidoc.pubman.appbase.FacesBean
English
 
Help Guide Disclaimer Contact us Login
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Conference Paper

Reference Counting in Library Design - Optionally and with Union-Find Optimization

MPS-Authors
http://pubman.mpdl.mpg.de/cone/persons/resource/persons44766

Kettner,  Lutz
Algorithms and Complexity, MPI for Informatics, Max Planck Society;

Locator
There are no locators available
Fulltext (public)
There are no public fulltexts available
Supplementary Material (public)
There is no public supplementary material available
Citation

Kettner, L. (2005). Reference Counting in Library Design - Optionally and with Union-Find Optimization. In Library-Centric Software Design (LCSD'05) (pp. 1-10). College Station, TX, USA: Department of Computer Science, Texas A&M University.


Cite as: http://hdl.handle.net/11858/00-001M-0000-000F-2781-0
Abstract
Reference counting has been used and described in abundance. We present novel ideas aimed at class implementations in library design: (1) In library design, generic classes can have variable size, such that an optimal decision for or against reference counting is not possible. We \emph{postpone} this decision to the place of class use. (2) In a context, where equality comparison for the case of equality is expensive, e.g., for exact algebraic number representations, we \emph{unify representations} whenever equality was detected, thus effectively caching equality tests. We explain an efficient implementation based on an union-find data structure. (3) Reference counting and \emph{polymorphic class hierarchies} can be combined reusing the pointer in the handle class for the polymorphism. A policy-based generic C\texttt{++} solution realizes all ideas. \emph{Standard} allocators manage all dynamic memory.