Design of high-performance C++ package for handling of multidimensional histograms

Abstract

This paper evaluates different object-oriented models and design techniques for development of high-performance numerical C++ software to handle multidimensional data. Two fundamental approaches are considered: the classical inheritance driven class hierarchy and the static polymorphism techniques. The inheritance driven approach exhibits intrinsic drawbacks, whilst static polymorphism techniques based on traits may significantly increase the efficiency and extensibility of the design. In the paper, a special attention is given to the analysis of efficiency and extensibility of the proposed solutions in ODBMS environment. As the result of this reaserch the Hist OOgrams Package has been developed at CERN as a part of LHC++ environment. The results of this study may be helpful in object-oriented design of other numerical packages.