Performance of Hardware Accelerated Particle Swarm Optimization with Digital Pheromones on Dissimilar Computing Platforms

Abstract

Programmable Graphics Processing Units (GPUs) have lately become promising means to perform scientific computations. When appropriately formulated, population based algorithms such as Particle Swarm Optimization (PSO) can leverage the data parallel architecture of GPUs dramatically improving the solution efficiency characteristics. Prior work by the authors demonstrated the feasibility for using GPUs for solving multidimensional optimization problems with digital pheromones in PSO using OpenGL Shading Language (GLSL). However, the programmability of GPUs in recent years fostered the development of a variety of programming languages making it challenging to select a computing language and use it consistently without the pitfall of being obsolete or unstable. This especially applies to design industries that aim at reducing investment and maintenance costs on high performance computing and training their designers to use such equipment. Although different GPU computing languages are available, some hardware specific languages are designed to rake in performance boosts when used with their host GPUs (e.g., Nvidia CUDA). On the other hand, a few are operating system specific (e.g., HLSL). A few are platform agnostic lending themselves to be used on a workstation with any CPU and a GPU (e.g., GLSL, OpenCL). This paper attempts to compare the performance of digital pheromone PSO when implemented on different GPU computing languages. Recommendations will be made on a viable platform for searching multi-dimensional design spaces. In other words, the paper aims to be a useful resource for designers aspiring for using GPUs in their optimization processes.