BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:FCCM Preview: More FLOPS or More Precision? Accuracy Parameterizab
 le Linear Equation Solvers for Model Predictive Control - Antonio Roldao (
 PhD@IC)
DTSTART:20090323T143000Z
DTEND:20090323T150000Z
UID:TALK17394@talks.cam.ac.uk
CONTACT:Dr George A Constantinides
DESCRIPTION:In this paper we exploit FPGA flexibility in the context\nof a
 ccelerating the solution of many small systems of linear equations\, a cen
 tral problem to model predictive control (MPC). The central observation ex
 ploited by this work is the distinction between accuracy (meaning the degr
 ee of correctness of a final computational result) and precision (meaning 
 the degree of correctness of each atomic computation). Using iterative met
 hods for solving linear systems\, one can obtain improved accuracy either 
 by running more iterations or by using more precise internal computations\
 , unlike direct methods\, where accuracy is only a function of operation p
 recision. Thus\, in iterative methods\, for a given accuracy requirement w
 e may conduct fewer iterations in a higher precision\, or more in a lower 
 precision. We argue that this suits FPGA architectures ideally\, as low pr
 ecision operations result in greater parallelism for any fixed area\nconst
 raint. We show that we may therefore optimize the performance by balancing
  iteration count and operation precision\, resulting in a several-fold spe
 ed improvement over a double-precision implementation\, but with the same 
 final result accuracy. Exploring this trade-off it is possible to provide 
 a speed-up of 26x on average\, 14x in the worst case and 36x in the best\,
  compared to a high-end CPU running at 3.0 GHz. This has the potential to 
 allow modern high performance\ncontrol techniques to be used in novel sett
 ings such as aircraft and diesel engines.
LOCATION:Mahanakorn Laboratory\, EEE
END:VEVENT
END:VCALENDAR