Performance Evaluation of Bitstring Representations in a Linear Genetic Programming Framework

TL;DR

This paper compares three bitstring implementations in C++ within a Linear Genetic Programming system, revealing platform-dependent performance differences and practical guidance for implementation choices.

Contribution

It provides a comparative analysis of bitstring representations in genetic programming, highlighting performance variations across platforms and offering selection guidance.

Findings

Custom implementation is fastest on Linux and Windows.

std::bitset performs best on macOS.

boost::dynamic_bitset is slower but flexible.

Abstract

Different bitstring representations can yield varying computational performance. This work compares three bitstring implementations in C++: std::bitset, boost::dynamic_bitset, and a custom direct implementation. Their performance is benchmarked in the context of concatenation within a Linear Genetic Programming system. Benchmarks were conducted on three platforms (macOS, Linux, and Windows MSYS2) to assess platform specific performance variations. The results show that the custom direct implementation delivers the fastest performance on Linux and Windows, while std::bitset performs best on macOS. Although consistently slower, boost::dynamic_bitset remains a viable and flexible option. These findings highlight the influence of compiler optimisations and system architecture on performance, providing practical guidance for selecting the optimal method based on platform and application requirements.