Author: Dann Corbit
Date: 17:59:45 05/28/04
Go up one level in this thread
On May 28, 2004 at 12:46:10, Anthony Cozzie wrote: >On May 28, 2004 at 12:37:31, Sune Fischer wrote: > >>On May 28, 2004 at 12:18:22, Anthony Cozzie wrote: >> >>>>These super rapid games are also great for stress testing, if nothing else. >>> >>>Definitely. I knew I had *zero* bugs in my threading implementation when Zappa >>>survived a 500 game 1 0 match with pondering against crafty without crashing or >>>losing on time :) >>> >>>anthony >> >>I'm still trying to find a good ponder implementation. >> >>I really want to avoid using peeknamedpipe so I use one thread for >>communication, one for pondering and N-threads for the search (currently N=1 >>always). >> >>I like to think of the search as being "the engine" and the communication is >>just something wrapped around it, like sort of an "engine interface". >> >>The problem for me is that pondering becomes something inbetween on its own >>thread. >>It really adds one level of complexity that just drives me nuts, e.g. when >>trying to stop pondering before or after the search has begun makes a >>difference. >> >>-S. > >Why? I have 2 threads: one does IO, and the other searches. I use select() to >determine if there is data, it times out every 20 ms. No idea if this works on >windows. There is no select() function on Windows except in Winsock, and that one only works on sockets, not on files.
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.