Author: Don Dailey
Date: 16:00:34 04/26/98
Go up one level in this thread
On April 26, 1998 at 18:37:15, Jan-Frode Myklebust wrote: >On April 26, 1998 at 17:54:24, Don Dailey wrote: >> >>Currently there is no plan for commercial products unless I do it >>with MIT which is not likely. However I am putting together a >>free parallel program. It will be very simple (along the lines >>of the simple public domain chess program currently available) but >>much better. >> >>I don't expect it to be competitive with Crafty or any of the >>current popular programs. If I'm lucky it'll be approximately >>as strong as GNU chess but that remains to be seen. >> >>I will not spend more than about 3 days on it so it certainly >>will not be fine tuned but will have all the standard stuff, >>hash tables, selectivity etc. The evaluation will be piece >>square table based and will be very primitive. I may add some dynamic >>evaluation if I can get to it quickly. Stuff like square of the >>pawn probable won't make it in though unfortunately. >> >>This will serve as a nice example of a parallel program and also show >>off our great language Cilk. To run it parallel, it will be much >>better to have a multiprocessor machine although one is not required. >>Of course it runs slower if you specify more processors than the machine >>actually has. Right now only Unix is supported for Cilk but the program >>will compile on any serial machine (as a serial program) if you have >>a standard ansi C compiler. But I can compile a version for any one >>who is interested. >> >>Here are the cilk platforms you can run on, they all must have Unix: >> >> 1) Dec Alpha's, >> 2) Intel pentiums >> 3) SGI machines >> 4) Sparc SMP's >> >>There is a version of Cilk for networks too but it's still under >>development and is not quite ready for chess. > >What do you mean by this? Will i be able to run it on the Universities >50++ Sun ultra's and get good performance from it, or will it need a >very fast connection between the machines?Sounds very interessting! I'm hoping this will be the case. We've had versions running on big distributed machines and they scale well. I can't say right now though how well it will work on a big slow network. The way global variables are handled is a big deal though. The hash tables implementation will have to avoid the shared memory (emulation) stuff and be reprogrammed with message passing for good performance. I'll let you know what we come up with. > >> >>By the way, we intend to port Cilk to Windows NT also. I do not >>know what the timetable is though, but probably pretty soon. >> >>I would like to set up a nice serial version for windows with a >>graphical interface if anyone shows any real interest. Does anyone >>know what is involved with kludging Xboard (the windows version) >>to work? I've done it on Unix but have no experience at all with >>Windows except for some minor system administration at the lab. >>Are there other user interfaces generally available (with protocol >>documentation?) > >Well, why not try to make a version for the Fritz interface (if the >protocol is freely available). Does anyone know about this? Is the interface protocol in the public domain or obtainable? Do I need library support? - Don
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.