BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:MRSC Preview: An FPGA-Based Floating Point Solver for Band Structu
 red Linear Equations - Antonio Roldao (PhD@IC)
DTSTART:20090323T140000Z
DTEND:20090323T143000Z
UID:TALK17393@talks.cam.ac.uk
CONTACT:Dr George A Constantinides
DESCRIPTION:Field Programmable Gate Arrays (FPGAs) have become an increasi
 ngly important technology in accelerating scientific computing (SC). A fre
 quent task in SC is the solution finding of systems of linear equations (L
 E). For finding such solution there are a number of well studied algorithm
 s. One algorithm that belongs to the class of iterative methods for solvin
 g LEs\, and has proven to be very ecient in hardware\, is the Conjugate G
 radient (CG) method [1]. In this work we present a hardware CG method whic
 h takes advantage of the banded structure present in many common problems.
  With the \nflexibility provided by FPGAs\, this implementation employs wi
 de-parallelization to convert the per iteration computation time for an or
 der n matrix with band width w from \\Theta(nw) clock cycles for a softwar
 e implementation to \\Theta(n) in hardware. It also explores deep-pipelini
 ng so that solutions to P problems are produced every \\Theta(n) cycles op
 posed to every \\Theta(Pnw) cycles in software.\n\n
LOCATION:Mahanakorn Laboratory\, EEE
END:VEVENT
END:VCALENDAR