Author: Robert Hyatt
Date: 06:19:57 09/30/99
Go up one level in this thread
On September 29, 1999 at 14:23:13, Dann Corbit wrote: >On September 29, 1999 at 05:46:09, Jari Huikari wrote: > >>On September 29, 1999 at 04:06:36, Will Singleton wrote: >> >>>On September 28, 1999 at 23:25:49, Dann Corbit wrote: >> >>>>I am going to use a paradigm that I don't think anyone else has even tried (but >>>>then again, I may be completely reinventing the wheel). >> >>>And, what might that be? >> >>I am interested too! I have also tried many things that others probably >>haven't. But so far it has been *removing* the wheels... :-) >Well, it has no commercial practicality, because it will suck if you don't have >a pile of processors. >Think of a chess program that starts not as main() but as a thread. Now, from >some position, I will start up a pile of threads -- each having a problem to do >analogous to a regular analysis of a position. For starters, I will have a >thread analyzing the current position -- which will be permanent throughout the >life of the move. Then, for each possible subsequent position, I will have a >thread examining each of those. Now, after a heartbeat interval, the threads >will check with a synchonization object and show what they have found so far. >The "dead ends" will be abandoned and the threads will look for new tasks (e.g. >for the most beneficial looking trails, they will examine subsequent positions >for those positions). Obviously, this will require a *PILE* of processors to >work well. However, on message passing machines or other non-SMP machines, it >should work like a charm. In other words, I could get the full horsepower of >one of those Silicon Graphics/Cray machines. It should also work well with a >pile of Hsu's chips. On a single CPU Intel machine, it will blow chunks and >die. Don't care, I don't want to sell it. There are a couple of problems... if you aren't going to use alpha/beta, and decide to try some sort of alternative like best first, then this kind of search might make sense. If you do stick with alpha/beta, the order in which you search things is critical, and which things you search first to establish an alpha/beta bound is also critical. Others have tried things like a normal engine with a cluster of machines working together, and also a stripped-down engine with a cluster of machines working to search a couple of extra plies tactically (Schaeffer did this in Phoenix, calling the stripped-down program 'minix'). It is an interesting idea although for anything but the longest of time controls network lag/jitter/latency will be a huge issue...
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.