Author: Robert Hyatt
Date: 14:32:04 09/27/01
Go up one level in this thread
On September 27, 2001 at 16:51:46, Uri Blass wrote: >On September 27, 2001 at 16:31:29, Robert Hyatt wrote: > >>On September 26, 2001 at 13:02:12, Robert Hyatt wrote: >> >>>On September 26, 2001 at 11:41:03, Vincent Diepeveen wrote: >>> >>>>On September 26, 2001 at 11:39:05, Vincent Diepeveen wrote: >>>> >>>>please also print for each depth the number of nodes you need! >>>> >>>>depths 12/13 with R=2 or R=3 you get within a few seconds or so.\ >>> >>>I don't get depth 12/13 in a few seconds. It is more like a few minutes. >>>What use is the node count? In chess you don't get to search until you >>>have searched X nodes. You get to search until you have used Y seconds... >>> >>> >>> >>>> >>>>All the 2.0 speedups i get here i never get within a few seconds. >>>>it's always levels which i can't realize in tournaments. >>> >>>I don't have that problem. I can get 1.9 in a few seconds or a few minutes >>>or a few hours. >> >> >>I assume this topic is "closed"? No more comments since I posted some real >>data to show that null-move has nothing to do with parallel search efficiency... > >No more comments does not mean that there is an agreement. > >I do not have an opinion about the speed improvement but I think that the right >test is not to search to a fixed depth but to search until you find a move that >you need to find. That is the same thing. There is no speed difference between the time at the point where you find a key move, vs using the time you finish that ply. A fixed depth is just a better way of controlling a test so that it finishes in some finite/fixed time. > >If the question is about long time control then you should use positions from >games when Crafty changes it's mind after a long time when you can compare the >time that it need to change it's mind with one processor and the time that it >needs to change it's mind with 2 processors. That time is _exactly_ proportional to the time to search to the same fixed depth. I can supply some logs if you want to see what I mean... IE it doesn't matter whether you take the time when the first PV is announced at depth=12, the time when the last PV is announced at depth 12, or the time when depth 12 is completely finished. The speedups are consistent. > >The positions do not need to be taken from a tactical test but from games when >Crafty changes it's mind after enough time always to the same move for >positional or tactical reasons. > >Uri I don't see what you are getting at. The question was "does a program that doesn't use null-move at all behave more or less efficiently when doing a parallel search than a program that does use null move?" It doesn't matter whether you time to a PV, or to the end of a fixed depth, when answering that question. The times are all proportional to each other...
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.