Research And Application Of Ray Tracing Technology For The Sunway Architectures

With the development of multimedia technology and CG technology,rendering engine has been widely used in movie animation,simulation and game effects.Mainstream rendering engine without exception attaches great importance to the important role of ray tracing algorithm in the rendering system,ray tracing is widely used for highly realistic rendering,can generate a true global lighting effect.However,this algorithm requires a large number of operations,causing slow in the current CPU architecture,a bottle of pre rendering,improves the film production costs,but also limit the application of the ray tracing algorithm is further more widely.Therefore,the acceleration of ray tracing has always been a problem worthy of studyThe GPU and Intel MIC ray tracing optimization work has a lot of effective results,in recent years,all using our self-developed domestic multi-core processing to build the Taihu supercomputer Shenwei light has become the world's fastest supercomputer,which uses on-chip heterogeneous manycore architecture provides a good platform as the acceleration of ray tracing.The architecture for on-chip integration of 64 core computing and a management control core contains four control core and 256 core computing a CPU,each management control core and 64 core Computing constitute a core group,flexible programming,suitable for the field of wells for rendering acceleration algorithm.In order to solve the problem of large amount of calculation and slow rendering speed in the process of real time rendering ray tracing,this paper presents a method of ray tracing acceleration based on domestic nuclear architecture.In the preprocessing stage,the four branch scene acceleration structure is constructed to adapt to the hardware architecture of domestic satellite.In the first stage of ray tracing,the ray tracing management control core control overall process,using the core strategy optimization of multi thread scheduling,scheduling of computing core light and scene tree intersection operation,to achieve control of nuclear and nuclear data asynchronous transmission,make full use of the main computational ability of nucleus;in the calculation of nuclear and light in the scene the tree intersection process presents a parallel algorithm for intersection,make full use of the nuclear production processing unit,the light and the scene tree 4 node parallel intersection to quantization operation,in order to speed up the process of intersection.The experimental results show that,compared with the CFU native mode,the proposed method can achieve 2 to 4 times the acceleration effect in the ray intersection stage,and the overall ray tracing process rendering speed is significantly improved.