Stockfish 4 PA_GTB

[echai]

Estimado Jeremy Bernstein
Estoy usando su versión GTB DD para el análisis infinito por la función de Persistent Hash.
Quisiera saber si se puede ampliar el número 32 por defecto de Persistent hash size y cuál sería el máximo conveniente teniendo 6 giga de memoria ram y 4096 de tablas hash
Igual pregunta para Persistent Depth
Muchísimas gracias.

Dear Jeremy Bernstein am using its GTB DD version for the infinite analysis by Persistent Hash function. I wonder if you can extend number 32 of Persistent default hash size and what would be the most suitable having 6 gig of memory ram and 4096 of hash tables the same question for Persistent Depth many thanks.

Dear Jeremy Bernstein Estoy using his version GTB DD for the infinite analysis for the function of Persistent Hash. It would like to be known if it is possible to extend the number 32 for shortcoming of Persistent hash size and what the suitable maximum would be having 6 jig of memory ram and 4096 of stage hash Equal question for Great Persistent Depth thank you.

Dear Jeremy Bernstein version I'm using his GTB DD to the infinite by the analysis of persistent hash function. I would like to know if you can increase the number 32 by default of Persistent hash size and what would be the maximum desirable taking 6 giga of ram and 4096 of hash tables Equal question for Persistent Depth Thank you.

Dear Jeremy Bernstein'm using version GTB DD for infinite analysis Persistent Hash function. I wonder if you can expand the number 32 Persistent default hash size and what would be the most convenient having 6 gig of ram and hash tables 4096 Same question for Persistent Depth Thank you.

[chri$]

come on Jeremy release anew version of stockfish with the GTB support.. last edition i have is the Sotckfish DD GTB.. stokfish comes out with new versionaily.. please atleast add a new compile for us who use Gaviota TB's only..(my Pc will not use syzygy for some reason.. only 345men work.6men wont work even with help from the houdini creator himself and various other friends who know alot about Tb's..

[Simon T]

come on Jeremy release anew version of stockfish with the GTB support.. last edition i have is the Sotckfish DD GTB.. stokfish comes out with new versionaily.. please atleast add a new compile for us who use Gaviota TB's only..(my Pc will not use syzygy for some reason.. only 345men work.6men wont work even with help from the houdini creator himself and various other friends who know alot about Tb's..

Perhaps your Syzygy tablebases corrupted. Did you check the md5sums?

[Jeremy Bernstein]

chri$ I appreciate your enthusiasm, but let me make this clear: no, I have neither the time nor interest to keep up with every commit on the Stockfish trunk. DD was, what, just 2 weeks ago? But I'm not a build server. I make new compiles as I have the opportunity, or when I see that a particular milestone has been reached, and that's just the way it is. You have the source code, you can fork my repository and merge from the upstream master just as well as I can and compile your own version if you need the latest and greatest all the time. Sorry.

Your syzygy problems sound like you need to run RdM's "tbcheck" program and figure out which file is corrupt.

jb

[chri$]

sorry jeremy my oost was intended with a bit of sarcasm as i was just giving you a good compliment, your compiles with the GTB feature have been a huge blessing for me since im not able to use the new syzygy tb's. i have ran a check of the md5 files and ive tried using just a few of 6men to see if sa file was corrupted.. i will try and redownload all of them again which took me almost 2 days to do on a 25 mbps upload connection and a 50 mbps download connection.. i will try that program to check for corruption.. where the 6men are for sure the problem.. my question is if u use just the 345men how much ELO gain do u get if u use the 6 men vs just the 3,4,5men syzygy tb's.. or if u use the syzygy tb's over not using them at all or using the gaviota tb's insted with the compile u made?

[Jeremy Bernstein]

Hey, sorry if that was a misunderstanding -- text communication always runs the risk... But still, no!

I gave up downloading the syzygy bases because the torrents in Germany were unbelievably slow (it was going to take somewhere between 3 months and 3 years according to uTorrent). I am now in the process of building them using MinGW on my 3770K/32GB PC -- I have about 70GB done so far, building around 16 hours/day for the last week. Should be done before the new year, I guess. So I haven't done anything with them yet, not even with the 3-4-5, which were done in a couple of hours.

I still use the Gaviotas at the moment with Stockfish (which I don't use much) but I'm interested in trying the syzygy bases since they're all the rage. My main engine (Komodo) doesn't currently support TBs, though, so I'm not suffering too much. I "accidentally" bought Houdini via a Chess Assistant/Aquarium update, so I now have a couple of engines (incl. Stockfish) which support the syzygy bases.

I still think that you should use the utilities on the Github site to verify all of the bases that you have. I can build them for you if you can't do that yourself.

jb

[KilgoreTrout]

come on Jeremy release anew version of stockfish with the GTB support.. last edition i have is the Sotckfish DD GTB.. stokfish comes out with new versionaily.. please atleast add a new compile for us who use Gaviota TB's only..(my Pc will not use syzygy for some reason.. only 345men work.6men wont work even with help from the houdini creator himself and various other friends who know alot about Tb's..

Are you on a 32 bit system? Syzygy 6-piece will only work with 64 bit:

http://talkchess.com/forum/viewtopic.php?t=50523

[KilgoreTrout]

BTW, I am running a tournament comparing "The Big 3" engines with TBs on and off, including the SF-DD-Gaviota compile that was kindly offered by Jeremy here. It will take months to complete, and get the error bars lower, but the tournament can be followed at this link as I update periodically:

http://talkchess.com/forum/viewtopic.php?t=50396

Latest results, SF-DD-Gaviota by Jeremy in #2, but of course the uncertainty is still large:

http://talkchess.com/forum/viewtopic.ph ... =&start=10

Program Elo + - Games Score Av.Op. Draws

1 Houdini_4_Pro_x64B-Syzygy : 39 31 30 216 56.2 % -4 56.9 %
2 Stockfish-DD-64-PA_Gaviota : 15 30 30 216 52.3 % -2 58.3 %
3 Houdini_4_Pro_x64B-Gaviota : 12 29 29 216 51.9 % -1 62.0 %
4 Komodo6-64bit : 10 30 30 216 51.6 % -1 57.9 %
5 Stockfish_DD_Syzygy_13112921 : 9 29 29 216 51.4 % -1 61.1 %
6 Stockfish_DD_64_modern : 7 30 29 216 51.2 % -1 59.7 %
7 Houdini_4_Pro_x64B-None : 4 30 30 216 50.7 % -0 57.9 %
8 Komodo-TCEC-64bit : 0 32 32 216 50.0 % 0 53.7 %
9 Houdini_3_Pro_x64-Gaviota : -32 29 29 216 44.9 % 4 60.2 %
10 Houdini_3_Pro-None : -65 31 31 216 39.8 % 7 54.6 %

[User923005]

Stockfish PA GTB is the breakfast of champions.

[echai]

Code:
Games Completed = 100 of 100 (Avg game length = 58.112 sec)
Settings = Gauntlet/64MB/30000ms+50ms/M 6500cp for 4 moves, D 150 moves/EPD:C:\Users\Paul\Documents\ChessBase\Testsets\EPD Test Suites\5000 8 move.epd(4962)
Time = 3016 sec elapsed, 0 sec remaining
1. Stockfish 151213IP 47.5/100 15-20-65 (L: m=0 t=0 i=0 a=20) (D: r=46 i=14 f=3 s=0 a=2) (tpm=423.6 d=16.37 nps=626701)
2. Stockfish PA_GTB DD 64 52.5/100 20-15-65 (L: m=0 t=0 i=0 a=15) (D: r=46 i=14 f=3 s=0 a=2) (tpm=438.9 d=16.36 nps=593099)
PGN: https://mega.co.nz/#!iRISQRoB!IacxAF...q_6YChCzSVa9_k
Paulied test