基于基元重启的OpenGL和CUDA图形渲染算法探索

Exploration of Graphics Rendering Algorithms Based on OpenGL Primitive Restart and CUDA

针对OpenGL渲染图形要多次访问缓存区的问题,提出一种OpenGL和CUDA混合编程的图形渲染算法来加速Perlin Kernel生成虚拟地形图。首先,通过OpenGL将缓存映射到CUDA内存空间,利用CUDA完成加速计算任务;然后,为几何图形设置开始和结束的位置标志,使用基元重启对图形进行组合;最后,对缓冲区对象进行渲染。实验结果表明,改进后的基于基元重启的混合算法在GTX 650GPU上的平均帧速率为960fps,帧速率提高6%,算法改进后渲染方法的执行效率提高了63倍。实验证实基元重启可以提高3D处理性能。

Aiming at the problem that the OpenGL rendering graphics should be accessed to the buffer zone multiple times,agraphics rendering algorithm based on OpenGL and CUDA mixed programming was proposed to accelerate the virtual terrain map generated by Kernel Perlin.First,the OpenGL is mapped to the CUDA cache memory to accelerate the completion of computing tasks.Then,the start and end position marks for the geometry were set by using primitive restart for the combination of graphics.Finally,the buffer object was rendered.The experimental results show that the average frame rate of the improved primitive restart algorithm in GTX 650 GPU is 960 fps,i.e.,the average frame rate increased by6%,and the execution efficiency of the improved algorithm after rendering method has increased 63 times.The experiments have confirmed that the primitive restart can improve the 3Dprocessing performance.