camera geometry in bevfusion has problem

[LIUHAO121]

https://github.com/NVIDIA-AI-IOT/Lidar_AI_Solution/blob/master/CUDA-BEVFusion/src/bevfusion/camera-geometry.cu#L147
wehn run this line geometry_out[_pid] = (coords.z * geometry_dim.z * geometry_dim.y + coords.x) * geometry_dim.x + coords.y;
coords.z will be 0 because voxel size, this line will be geometry_out[_pid] = coords.x * geometry_dim.x + coords.y;
why not coords.x * geometry_dim.y + coords.y or coords.y * geometry_dim.x + coords.x ?