BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:How do I know my GPU or multicore floating point computations are 
 correct? - Miriam Leeser - Northeastern University
DTSTART:20110601T120000Z
DTEND:20110601T130000Z
UID:TALK31368@talks.cam.ac.uk
CONTACT:William Denman
DESCRIPTION:GPU and multicore architectures are often used to accelerate s
 cientific applications that rely on floating point computation.  Current t
 echniques for verifying the correctness of these implementations are quite
  primitive.  Scientists expect their floating point code to produce differ
 ent results on different architectures\, and as a result do not test their
  code thoroughly.  \n\nThe talk will consider two case studies:  a paralle
 l implementation of Pi and a real world medical application used in cancer
  diagnosis.  Through these case studies\, the sources of differences in CP
 U and GPU codes will be identified.  Issues of correctness\, including con
 trol issues such as thread synchronization as well as issues of floating p
 oint correctness will be discussed.  I will give a brief survey of tools a
 vailable for addressing these issues and discuss current research directio
 ns in identifying bugs in GPU and multicore implementations.    \n
LOCATION:Computer Laboratory\, William Gates Building\, Room SS03
END:VEVENT
END:VCALENDAR