View Single Post
Old 2011-04-05, 17:38   #25
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

3×2,083 Posts
Default

Quote:
Originally Posted by Rick Reynolds View Post
Don't suppose there's any 32 bit cuda enabled llr versions running around for this rally or is that even possible here?

Rick
To further elaborate on what Ralf explained: we do have a CUDA LLR program, though for numbers of this size it's not much use (as it's only about the speed of one CPU core on a modern GPU such as a GTX 460). It is somewhat faster for much larger numbers, i.e. PSP and SoB, but not for what NPLB's testing right now. As a point of reference, the numbers we're doing in this rally are in the vicnity of n=850K; Gary has a GPU that he's currently running on LLRnet port 3500, for another one of our efforts in the vicinity of n=1.3M, and it's about twice the speed of a CPU core.

I do have a binary on hand for llrCUDA, but it's for 64-bit Linux. What operating system are you looking to run this on? There have been some 64-bit Windows binaries posted in the forum (see here for the latest), but I don't believe I've seen any for 32-bit Windows. For 32-bit Linux, however, it should be pretty easy to compile the latest source as long as you have the CUDA toolkit installed.
mdettweiler is offline   Reply With Quote