Demonstration of GPGPU-Accelerated Computational Fluid Dynamic Calculations

Problems in computational mechanics involve higher order nonlinear differential equations with complex boundary conditions, which are difficult to solve analytically and need numerical methods to predict the approximate solution. A large number of mesh points are utilized for better accuracy of the numerical technique which results in storage and operations of a large amount of data. It is of utmost importance that the time taken to perform these calculations is reduced to a realizable scale. General purpose graphical processing units (GPGPUs) provide high number of floating point operations per second (FLOPS) and potentially offer the most efficient architecture to carry out large-scale calculations in computational mechanics. In the present work, an attempt has been made to reduce the computational time for obtaining numerical solution of heat transfer by conduction, laminar flow in a rectangular channel, lid-driven cavity, and flow past square cylinder by programming GPGPUs using compute unified device architecture (CUDA), while maintaining overall second-order accuracy.