论文部分内容阅读
为满足天文观测及激光通信等领域的1000单元级以下的自适应光学(AO)系统的高速实时处理的要求,提出一种单板上实现高速实时计算的通用化技术方案。采用多核CPU实时计算波前,结合高速图像采集卡,以满足系统高帧频的需求;计算机系统采用Xenomai实时操作系统及Linux操作系统,以满足AO系统低延迟的需求;实现了在单台计算机上集成实时处理与监控的紧凑型处理系统。采用向量指令优化及多核并行运算,在6核的计算平台上获取了峰值每秒510亿次浮点运算的计算能力,102 GB/s的通信带宽。仿真结果表明可以对1145个驱动器,949个子孔径的系统进行2000 Hz的处理,处理延迟低于240μs;通过修改参数后,应用于光通信的137单元AO系统中,校正后斯特列尔比为0.61。结果表明该处理方案能满足1145单元以下AO系统2000 Hz实时处理的需要,并具有通用性。
In order to meet the requirements of high-speed real-time processing of adaptive optics (AO) systems below 1000 level in astronomical observation and laser communications, a generalized technical scheme for realizing high-speed real-time computation on a single board is proposed. The multi-core CPU calculates the wavefront in real time and combines the high-speed image acquisition card to meet the high frame rate requirements of the system. The computer system uses the Xenomai real-time operating system and the Linux operating system to satisfy the low latency requirement of the AO system. Compact processing system with integrated real-time processing and monitoring. With vector instruction optimization and multi-core parallel computing, the computing power of 51 billion floating-point operations per second and 102 GB / s communication bandwidth are obtained on a 6-core computing platform. The simulation results show that it is possible to process 2000 Hz with 1145 drivers and 949 sub-aperture systems with the processing delay less than 240μs. After the parameters are modified, the 137-unit AO system used in optical communication is calibrated with the Striglian ratio 0.61. The results show that this scheme can meet the need of 2000 Hz real-time processing of AO system below 1145 units, and has the versatility.