[parsec-users] Parsec GPU Implementations?

Jim Dempsey jim at quickthreadprogramming.com
Wed Oct 19 08:50:17 EDT 2011

Hi Manish,
While I haven't explored GPGPU implementation of the PARSEC benchmarks, I
will likely use them to explore these benchmarks using the Intel Many
Integrated Core (MIC) Knights Corner (as soon as I can get one). I have
experimented using the ATI FireStream two years ago and much has happened
since then. At that time double precision was not fully implemented (e.g.
lost precision on division and no trig functions), leading to lesser
accuracy and performance than increasing CPU thread count. I anticipate the
Intel Knights Corner will resolve most of the remaining issues. For my own
interests it is accuracy + speed as opposed to speed alone.
Jim Dempsey


From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Manish
Sent: Wednesday, October 19, 2011 12:14 AM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] Parsec GPU Implementations?

Hi All

During the ISCA tutorial Parsec developers were soliciting GPU
implementations of the various benchmarks. From my search (Matt Sinclair's
work) I have found evidence of blackscholes, fluidanimate and streamcluster
(swaptions as well but slower than CPU) being ported onto GPU's. Have the
parsec developers received any other successful implementations? Does any
one know of any other pointers to GPU implementations of other benchmarks?


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20111019/940abe32/attachment.htm>

More information about the parsec-users mailing list