A high-performance FPGA platform for adaptive optics real-time control

Adaptive Optics Real-Time Control systems for next generation ground-based telescopes demand significantly higher processing power, memory bandwidth and I/O capacity on the hardware platform than those for existing control systems. We present a FPGA based high-performance computing platform that is developed at Dominion Radio Astrophysical Observatory and is very suitable for the applications of Adaptive Optics Real-Time Control systems. With maximum of 16 computing blades, 110 TeraMAC/s processing power, 1.8Terabyte/s memory bandwidth and 19.5 Terabit/s I/O capacity, this ATCA architecture platform has enough capacity to perform pixel processing, tomographic wave-front reconstruction and deformable mirror fitting for first and second generation AO systems on 30+-meter class telescopes. As an example, we demonstrate that with only one computing blade, the platform can handle the real time tomography needs of NFIRAOS, the Thirty-Meter Telescope first light facility Multi-Conjugate Adaptive Optics system. The High- Performance FPGA platform is integrated with Board Software Development Kit to provide a complete and fully tested set of interfaces to access the hardware resources. Therefore the firmware development can be focused on unique, userspecific applications.