Subdivision Meshes in GPU 20075026

Introduction The most of objects in our real life has smooth shape. The subdivision meshes are developed for representing the characters and the objects naturally in animations and games. The subdivision meshes are methods of representing the smooth surfaces using iterative operation with a few number of data (control meshes). The graphics processing unit (GPU) in nowadays has more flexible structure and programmability while the traditional graphic processors with the fixed pipeline were optimized for rendering the image to the frame buffer. These programmable elements in GPU are called the programmable shaders. Thanks to the programmable shaders, GPU can operate the complex calculation or flow control like CPU. GPU is suitable for processing independent and parallel data due to its structural characteristic. The various evaluations of subdivision meshes on GPU have been researched since the algorithm of subdivision meshes is matched well with GPU structure. Many people want to use the GPU for solving general problems – physical simulation, calculations, etc. – because the GPU is very powerful parallel processing machine. The general purpose graphics processing unit (GPGPU) fields handle those problems. Even though GPU has a plentiful of programmable functionalities, the GPU has a limitation of data feedback – reusing the processed data – since GPU is unidirectional processing unit for rendering image. In this research, we survey the structural features of GPU and find the most efficient memory handling for resolving the bottle-neck of memory transferring during evaluating of the subdivision meshes.