BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Performance-portable Programming Abstraction for Image Processing 
 - Richard Membarth\, University of Erlangen-Nuremberg (currently ARM inter
 n)
DTSTART:20101022T150000Z
DTEND:20101022T160000Z
UID:TALK27514@talks.cam.ac.uk
CONTACT:Alan Mycroft
DESCRIPTION:We present a framework for representing image processing kerne
 ls based on decoupled access/execute metadata\, which allow the programmer
  to specify both execution constraints and memory access pattern of a kern
 el.\nThe framework performs source-to-source translation of kernels expres
 sed in high-level framework-specific C++ classes into low-level CUDA or Op
 enCL code with effective device-dependent optimizations such as global mem
 ory padding for avoiding partition conflicts.\nWe evaluate the framework o
 n two kernels on GPU and CPU systems\, and concur with previous work that 
 access/execute metadata is a suitable performance-portable abstraction for
  image processing.\n
LOCATION:GS15\, Computer Laboratory
END:VEVENT
END:VCALENDAR