Author: Heiner Marxen
Date: 10:52:20 09/19/02
Go up one level in this thread
On September 18, 2002 at 15:57:58, Vincent Lejeune wrote: >On September 18, 2002 at 14:42:57, Heiner Marxen wrote: > >>On September 18, 2002 at 14:12:18, Vincent Lejeune wrote: >> >>>On September 18, 2002 at 13:56:21, Heiner Marxen wrote: >>> >>>>On September 18, 2002 at 13:17:26, F. Huber wrote: >>>> >>>>>On September 18, 2002 at 01:12:15, Jouni Uski wrote: >>>>> >>>>>>I have tested UCI version and found, that program is unbelievable fast. I have >>>>>>tested some EGTB positions to verify solution with Nalimov's! And they were >>>>>>same. >>>>>>But one question: is it possible to see any info about search depth and/or node >>>>>>speed? Or should I only wait may be hours before anything is displayed? >>>>>> >>>>>>Jouni >>>>> >>>>>Hi Jouni, >>>>> >>>>>that´s really one problem, I´m also not very happy about with my ChestUCI. >>>>>But the original ´Chest´ does NOT output ANY informations while thinking, >>>>>and so also my UCI-interface has absolutely no chance to inform the GUI about >>>>>the progress of the searching :-( >>>>> >>>>>If you really have no idea of ´how many moves to mate´, maybe it could help >>>>>to start with a moderate number of ´SearchDepth´ (e.g. 5, 6,...) and increase >>>>>this number by one whenever no mate has been found. So you could at least >>>>>get ´some´ idea, how long the next calculation step would need. >>>>> >>>>>Maybe in the next version Heiner Marxen will implement ´some´ output info >>>>>while thinking - at least the actual search depth would IMO not slow down >>>>>the reallly fantastic seach speed. >>>> >>>>I'm listening... "I'll do my very best". >>>> >>>>Unfortunately, the next version of Chest is delayed, but there is some chance, >>>>that I offer a slightly extended version 3.19b, better suitable for ChestUCI. >>>>I would announce that here, of course. >>>> >>>>Cheers, >>>>Heiner >>> >>>It would be great if you can make a .pgn file output with variations and >>>sub-variations ! >> >>That point is much further down in my TODO list. >>Besides the numerous details of PGN (Chest does know EPD, but not yet PGN), >>one problem is, that a complete solution tree can become _SO_HUGE_ >>that it would not even fit onto any disk. >>So, how to select, what to include, and what not? >>Clearly, the PV shall be there. But exactly what else? >> >>>Thanks for your marvelous mate solver ! ;o) >> >>:-)) >> >>Cheers, >>Heiner > >May be you can simply stop writing file when size >100 Kbytes or when output >have more then 100 line or so ... That would put the most uninteresting variants into the output, and leave out the more interesting ones. I already knew that we want a size limit on the output. ;-) The problem is to _select_ the contents. Otherwise there is no point in printing more than just the PVs, at least when the complete output would be much too large. Ideas for selecting the most valuable/interesting parts? I have already thought about "threats" and omitting lines that do not fight against a threat... but that is still somewhat vague. What exactly is a threat? Ok, I got that (NULL move for opponent followed by mate). Threats are not yet known to Chest, they need to be computed. That can be expensive. Well, we could reduce the depth for that. Then, what exactly does that mean: "does not fight against a threat". Here I'm not sure how to program that. Still, even with "threat" logic the solution trees can be _huge_. Now what? Cheers, Heiner
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.