Performance Optimization Scheme of Vehicle-Mounted Computing Platform Based on OpenGL

In order to overcome the problems of insufficient computing power in complex computing scenarios due to hardware limitations of vehicle-mounted embedded devices and too low rendering frame rate in multi-element rendering, a new comprehensive optimization method is proposed. First, this method converts the large number of floating-point numbers that originally needed to be calculated in the CPU into texture data, also, performs parallel calculations by OpenGL in the GPU, which solves the efficiency problem of the on-board system when a large number of floating-point is calculated. At the same time, this method constructs a directed acyclic graph based on the FrameGraph architecture, which can effectively improve the performance of rendering a large number of elements. On the basic of the above, it can improve the calculation speed of forward computing scenarios that require deep learning such as face recognition and pedestrian detection, and solve the efficiency problem of rendering a large number of elements in the vehicle large screen auxiliary display. The experimental results show that the proposed method can reduce the computing time by 33%, and the rendering frame time by 14%.