Hardware Implementation of Spherical Diffuse Reflection in Ray-Casting Using High Level-Synthesis

Game applications require high graphics, and it is not suitable for mobile devices which is battery-powered because of heavy processing load. Mobile terminals equipped with FPGA, which are dynamically reconfigurable hardware, can take advantage of the high processing performance and low power consumption of FPGA, thus helping to solve these problems. However, this system tends to have high development costs due to the need of circuit data for FPGA's operation. To avoid this problem, we make use of High-Level Synthesis, HLS. HLS is a tool that automatically generate hardware from software program, but it is not panacea. If there is no consideration of hardware structure, it sometimes generates inferior hardware. We are therefore developing a HLS-oriented program particularly focusing on ray-casting, which is utilized in the creation of high-graphic game applications, and have implemented spherical diffuse reflection which is one of the processes included in ray-casting. In this paper, we implemented fixed-point conversion to make numerous vector operations used in ray-casting more hardware-friendly. The experiment evaluates that the circuit size, execution time, and power efficiency. The results show that the fixed-point method contributes significantly to the reduction of the circuit size and the created hardware performs satisfactorily against the software.