21 July 2014 Benchmarking hardware architecture candidates for the NFIRAOS real-time controller
Author Affiliations +
As a part of the trade study for the Narrow Field Infrared Adaptive Optics System, the adaptive optics system for the Thirty Meter Telescope, we investigated the feasibility of performing real-time control computation using a Linux operating system and Intel Xeon E5 CPUs. We also investigated a Xeon Phi based architecture which allows higher levels of parallelism. This paper summarizes both the CPU based real-time controller architecture and the Xeon Phi based RTC. The Intel Xeon E5 CPU solution meets the requirements and performs the computation for one AO cycle in an average of 767 microseconds. The Xeon Phi solution did not meet the 1200 microsecond time requirement and also suffered from unpredictable execution times. More detailed benchmark results are reported for both architectures.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Malcolm Smith, Malcolm Smith, Dan Kerley, Dan Kerley, Glen Herriot, Glen Herriot, Jean-Pierre Véran, Jean-Pierre Véran, "Benchmarking hardware architecture candidates for the NFIRAOS real-time controller", Proc. SPIE 9148, Adaptive Optics Systems IV, 91484K (21 July 2014); doi: 10.1117/12.2057345; https://doi.org/10.1117/12.2057345

Back to Top