The paper compares three APIs for the implementation of cloth modelling on modern graphics processor units (GPU): OpenGL plus GLSL, NVIDIA CUDA and OpenCL. They are compared by programming features, platform and device portability, and performance for the purpose of dynamic cloth simulation. Results about performance are given and conclusions are drawn about use cases.

This paper addresses the difficulties of teaching introductory programming to students in the field of information technologies and computer science. It reviews several existing approaches to make learning programming more attractive: using games or programmatically controlled hardware devices. The author believes that programming moving virtual actors motivates most of the students to learn programming. Moreover, seeing moving visual objects on the computer screen makes it easier for the students to perceive the mistakes in their code and correct them. However, writing such programs without any additional tools is not a task for a novice programmer. That is why this paper proposes an approach for teaching programming for IT professionals using a new game library, specially developed for training, which controls virtual models moving in a virtual environment. The system uses appealing 3D graphics to attract attention. Several sample programs, illustrating main features of this approach, are presented. Screenshots of the sample programs and results of teaching a pilot group of students are given at the end of the paper. Initial results show that the students enjoyed the course, most of them liked the game library approach better than the traditional one and do recommend using it in the future. More than half of the trainees improved their results.

This paper describes several techniques for accelerating a virtual try-on garment simulation on a mobile device (smartphone or tablet) using parallel computing on a multicore CPU, GPU computing or both depending on the mobile hardware. The system exploits a mass-spring cloth model with velocity modification approach to overcome the super-elasticity. The simulation starts from flat garment pattern meshes positioned around a 3D human body, then seaming forces are applied on the edges of the panels until the garment is seamed and several cloth draping steps are performed in the end. The cloth-body collision detection and response algorithm is based on image-space interference tests and the cloth-cloth collision detection uses entirely GPU based approach on the newer hardware or recursive parallel algorithm on the CPU. As the results section shows the average time of dressing a virtual body with a garment on a modern smart phone supporting OpenGL ES2.0 is 2 seconds and on a tablet supporting OpenGL ES3.0 or 3.1 is less than one second.