GPU Accelerated Rotation About an Arbitrary Axis

Abstract

The three-dimensional (3D) rotation about any axis is essential in diverse applications and fields, particularly in computer graphics. This paper focuses on accelerating the operation of this transform using GPU in a real-time environment. This special type of rotation is complicated compared to rotation about the conventional axes due to having many matrix operations, so accelerating such a transform with parallel techniques is an important issue to reduce the execution time that is important to ensure the realistic view of 3D animation scene. In addition to that, concatenating these many operations in a single rotation matrix also gives a significant reduction in computation time required to perform the rotation. The rotation transform is applied to complex models with hundreds or even millions of vertices, so standard 3D objects with different resolutions are used for testing the rotation about a selected axis that created interactively using LabVIEW and Visual Studio software environments. The experimental results showed the significant speedup on CUDA/C++ compared to LabVIEW computations for the same model complexity.