The Research And Development Of The Point Based Global Illumination Algorithm On The Intel MIC Architecture

With the development of Computer Graphics, the multimedia field has been promoted greatly, and takes profound changes into people’s lives. The High Performance Computing’s rapid growth takes a chance to Computer Graphics development. The research on computer graphics could be turned from the simple illumination model to the physics-based illumination model. The 3D rendering technology is becoming more and more important in multimedia industry, and people start to continuously pursue higher-quality and more realistic render effects. Global illumination effect as an important part of photorealistic rendering, becoming the focus of realistic rendering development. Global illumination effect refers to the process of three-dimensional scene rendering in computer, it need the influence of direct and indirect light. The three-dimensional visual scene is more realistic and richer detail from the calculation of global illumination effect. The algorithms to calculate global illumination effects can involve huge computational process, the requirements of time and hardware space for realistic rendering has gradually become the restriction of the art products. To speed up rendering work, shorten the production cycle, becoming a key point in the rendering technique.On present, Ray Tracing, Photon Mapping and Point Based Global Illumination (PBGI for the short) are the most studied parts of the global illumination algorithm.PBGI is mainly applied film and animation industries, and noiselessly simulating ambient occlusion, color bleeding, soft shadow and other more effects is its greatest advantage.At the end of 2012, Intel launched Xeon Phi as the coprocessor of Xeon series, which based on Many Integrated Core Architecture(MIC for short). MIC has more than 50 computer cores that based on X86 architecture, and each core is equipped with a very wide 512 bit vector processor unit(VPU for short), with this it can process the operation of 16 single-precision floating-point numbers at the same time. At the same time, the coprocessor with MIC architecture based X86 architecture which equipped quality instruction set and will bring programmer great convenience.So far, multiple supercomputers including Tianhe-2 have used the architecture which introduced CPU and MIC coprocessor. To develop PBGI algorithm effectively ran on the MIC processor has great significance.This paper researches on the problem that huge calculation involved in traversing point cloud hierarchal tree causes slow rendering speed, and presents a kind of PBGI accelerating technology based on Intel MIC. By optimizing point cloud hierarchal tree, using asynchronous data transmission, opening up memory space in advance on MIC and modifying algorithm to use Xeon Phi coprocessor vector computing units, PBGI algorithm speeds up remarkably on MIC.The background of this paper is the national 863 project’Development and application of realistic animation rendering system’, and the PBGI algorithm is integrated in realistic rendering engine-RenderWing, which is independently studied by Shandong University. This paper discusses that PBGI algorithm is able to run effectively on Intel MIC coprocessor by modifying it, and passes the performance test on super computer such as Tianhe-2. In the end, this paper summarizes inadequacies and some aspects which need to improve of vectorization of PBGI algorithm, and prospects of the future research.