3D Model Compression with Support of Parallel Processing on the GPU
A. V. Nikolaev, Vladimir Alexandrovich Frolov, Irina Gennadievna Ryzhova · Programming and Computer Software · 2022
Abstract In this paper, we propose a method for 3D model compression with fast GPU-accelerated decompression. This method applies on-the-fly decompression directly in the process of rasterization or ray tracing and allows us to fit three to eight times more geometry into the same volume of GPU memory. For rasterization, we implement two variants of decompression: via geometry and via mesh shaders. For ray tracing, we use hardware acceleration with the Vulkan VK_KHR_ray_tracing_pipeline extension and propose a BVH tree leaf caching technique, which speeds up rendering nearly twofold. The conclusions made in the process of performance evaluation of the proposed method can be useful to design hardware support for 3D model compression on future GPUs, because we ran into some hardware limitations of existing GPUs.