The Performance Improvement of Octave in Solving Various Problems Through Multithreading Technology

Abstract

GNU Octave is a widely used open-source scientific computing environment that provides a high-level programming language for numerical computation, data analysis, and algorithm development. However, Octave's single-threaded execution model limits its performance for computationally intensive tasks common in scientific research, including AI model training, parameter sweeping for differential equations, large-scale statistical data analysis, and finite element simulations.

This paper presents octave_boost, an open-source package that integrates the Boost C++ Libraries into GNU Octave, with a particular focus on the Boost.Thread module to provide multi-threading capabilities. Through a set of carefully designed C++ loadable functions, octave_boost enables parallel execution of Octave expressions, function evaluations, and script files using boost::thread, transparently managed by the Octave interpreter. The package also integrates Boost.Chrono for nanosecond-precision time measurement, Boost.Accumulators for online statistical computation, Boost.Date_Time for calendar arithmetic, and Boost.PropertyTree for configuration format conversion.

We conduct comprehensive performance benchmarks across four representative scientific computing scenarios: AI model training checkpoint recording, ordinary differential equation (ODE) parameter sweeping, large dataset statistical analysis, and finite element analysis visualization. The experimental results demonstrate that octave_boost's multi-threading capabilities achieve substantial speedups—up to 5.8× on a 6-core processor for embarrassingly parallel workloads—while maintaining nanosecond-level timing precision for performance measurement and checkpoint synchronization.

The octave_boost package is released under the GNU General Public License v3.0 and is available at https://github.com/CNOCTAVE/octave_boost, providing researchers and engineers with a practical and accessible solution for accelerating Octave-based scientific computations through multi-threading.