Author: Robert Hyatt
Date: 07:36:55 09/18/01
Go up one level in this thread
On September 18, 2001 at 07:34:11, David Rasmussen wrote: >On September 18, 2001 at 02:16:32, Gian-Carlo Pascutto wrote: > >>On September 18, 2001 at 01:11:34, David Rasmussen wrote: >> >>>What about OpenMP, MPI and all that? >> >>What about them? They're message-passing libraries/interfaces. >> >>APHID is a ready-and-done parallel game-playing lib to plug >>into your program. >> >>-- >>GCP > >Yes, but I still have to do some coding myself. On the SUN compilers, >auto-parallelization can be performed. It could be interesting to see how this >would compare to explicit SMP code. I could compile a "single processor" version >of Crafty with -autopar on the SUN compiler, to see the difference. Obviously it >won't be as good as explicit SMP, but it might still give something. My own >chessengine isn't SMP yet, so this would be an easy way to take advantage of the >96 processesors in largest machine at my university. Unfortunately, -autopar >only works with their C and Fortran compilers, not the C++ compiler. Dang... For a tree search, no automatic parallelizing compiler on the planet will do a thing for you. You will be lucky to get a 1% speedup... I've done this a few times for fun, using various forms of the KAP compiler (such as the one available for the Cray, etc.) It is simply not up to a recursive algorithm.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.