Author: Bob Durrett
Date: 11:16:29 02/03/04
Go up one level in this thread
On February 03, 2004 at 13:59:54, Matthias Gemuh wrote: >On February 03, 2004 at 12:49:53, William Penn wrote: > >>On February 03, 2004 at 11:02:05, Bob Durrett wrote: >> >>>On February 03, 2004 at 08:09:52, William Penn wrote: >>> >>>>On February 03, 2004 at 04:32:35, Matthias Gemuh wrote: >>>> >>>>> >>>>> >>>>>Zied Haddad seems to be unable to connect to CCC from Tunesia. >>>>>He seeks to know whether others have encountered a Shredder 8 problem he has: >>>>> >>>>>after playing many games in a row without clearing hash tables, Shredder 8 >>>>>starts messing and missing even mate in 1, till hash tables are cleared. >>>>> >>>>>[http://f50.parsimony.net/forum200321/messages/5855.htm] >>>>> >>>>>/Matthias >>>>>(message posted on Zied's request) >>>> >>>>You may be right. I'm not sure yet, but I will start clearing the hash tables >>>>before each analysis and see if it makes a difference. Note that I only run >>>>Shredder 8 in infinite analysis mode for a long time to analyze positions. >>> >>><snip> >>>> >>>>WP >>> >>>(1) I read somewhere that it's always a good idea to clear the hashtables >>>before starting a new analysis but do not recall the reason. Nevertheless, I >>>now always hit that "clear hashtable" button whenever I see it. >>> >>>(2) The PV for Shredder is total garbage except for the first move and this is >>>a known bug in Shredder, acknowledged by it's author. It hopefully will be >>>fixed one of these days. The sooner the better. >>> >>>(3) With some engines, I notice another bug which may or may not be related to >>>this one. While doing user-guided infinite analysis, sometimes the nps number >>>drops to a tenth of it's normal value. When that happens I cuss a little and >>>then save the position, turn off the engine, and then restart it, clearing the >>>hashtable. Then the nps goes back to a reasonable value. I do not recall which >>>engine does this, but it has to be either Crafty, Shredder, Fritz, Hiarcs, >>>Tiger, or Junior since those are the engines I use as CB8 analysis engines and >>>for overnight automated game analyses. >>> >>>(4) I still have not learned how to use Chessmaster for analysis, perhaps >>>because I'm getting overly senile. It's a pity that The King is not available >>>as a CB8 analysis engine because if it were I would use it often. >>> >>>Bob D. >> >>Thanks for the info. Can you tell me what "PV" means? I see it mentioned all the >>time but have never run across the definition. >> >>As for nps dropping drastically to 1/10th of it's normal value... That can >>happen in endgames with any engine due to heavy tablebase access, although it's >>not usually so drastic. You can tell if that's the case because of constant hard >>drive access (the hard drive activity light stays on). I've also seen some >>significant dropping of nps with the Shredder 8 Gambit engine, but again not so >>drastic, perhaps to 1/3rd of normal value. I haven't seen that happen with the >>native Shredder 8 engine, which is usually fairly constant speed in a particular >>position. There's also the possibility for the counter in some older engines to >>wrap back around to zero when it gets to a certain value, a large number which I >>don't recall offhand. That causes a big drop in nps, but it's only artifical - >>the speed hasn't actually slowed down. >>WP > > > >A Principal Variation is (from your posting) this thing : >28.Kf2 f5 29.Ke1 a4 30.Ke2 Kf7 31.g4 fxg4 32.Rc4 dxc4 33.b3 axb3 34.Kd2 bxa2 > >A dramatical drop in nps during long analysis is often due to counter overflow, >especially when programmer mistakenly uses counters of wrong type >(e.g. 32 bit counters instead of 64 bit) > >/Matthias. Thanks, Matthias, for making me feel better. I feared it was "stupid operator error" of the senile type. : ) Generally, we are back to the issue of whether or not commercial program producers REALLY care about the users. Sometimes I think users are merely a source of money. I would like to see flawless and meaningful outputs sent to the user always without exception. I guess it sometimes is not a real priority for the producer [not necessarily the programmer]. Bob D.
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.