Page 1 of 1
5-1 EGTB Status. What is needed to make them useable?
Posted: Tue Jun 19, 2007 6:59 pm
by ernest
Can somebody give the actual status of the 5-1 EGTBs ?
Is there a prospect that some day, they may be used in the same way as the other 6-man Nalimovs, in chess programs (that is, either during the search, or at least as a static database)?
My understanding is the following:
1. the existing Nalimov code does not allow for use of the 5-1 EGTBs in a search program. Thus Nalimov's code has to be modified...
2. Martin Kreuzer completed the generation of the 5-1 EGTB. How can they be used at present?
Generally, what can be planned?
Thanks for this layman's clarification...
Posted: Thu Jun 21, 2007 2:25 am
by jkominek
Hi Ernest,
Way back in early January, still feeling the liberation of Christmas holidays, I made headway in the task of extending tbgen for this purpose. In a spate of New Years resolutions of
Great Progress This Year I had, well, uh, visages of great progress -- of the generation code churning away for months on end, of emule-ing files for others to verify, of generating 5-1 stats that "vb4" would so much like to have hands on. Then came February. Poof.
(My "all-day and all-night job," is being a computer science PhD student at a prestigious university. "Prestigious" means you get worked like a dog for someone else's glory, for an interminable amount of time. Until your funding runs dry. Then you get kicked out. Details at
http://www.phdcomics.com)
I see some relief in sight, round about early July. So, events willing, expect another posting then. Maybe I'll have something concrete.
In the interim, I'm open to pledges of donated computing power. Sucking energy out of our compute cluster as school would not be kosher, and my personal machines are showing wrinkles around the corners of their eyes (just like me). I haven't been on this board much in the past several months, but, having announced my intention here, will check back for any takers.
john
Progressing 5-1 EGT generation
Posted: Fri Jun 22, 2007 9:39 pm
by guyhaw
Interesting that you have done something with TBGEN. Marc Bourzutschky would perhaps be interested in that. Certainly, one cannot have Nalimov 5-1 EGTs without an extension to TBGEN.
A way forward is to convert the FEG 5-1 to Nalimov format: again, this road leads back to Marc's conversion routine which may (or may not) begeneric enough to deal wtih this.
My interest hasn't changed: example maxDTx positions (DTC, DTM, DTZ), and zugzwangs (there are a few) would be good - if only to push the boundaries of completeness.
g
Posted: Fri Jun 22, 2007 11:34 pm
by ZeroOne
jkominek wrote:In the interim, I'm open to pledges of donated computing power.
I could donate some computing power... I have even just finished downloading all 5-men Nalimov bases. I just need a program and some instructions. I'm already running BOINC continuously anyway so I might as well run this thing.
Edit/Add: I'm running Windows XP.
Posted: Sat Jun 23, 2007 3:53 pm
by jkominek
Guy - Yes, I now recall your interest in extrema positions (maxDTMx and zugs) as being best for spot checks. Will co-ordinate with you on that when the time comes.
Ernest - Thank you. Some thought needs to go into the problem of distributed EGTB computation since, unlike SETI or protein folding or prime number searches, it is very heavy on the data requirements. 5-1 is a nice starting point since the tables are relatively small and the dependency graph less expansive. Also, be aware that tbgen has high memory demands. For some of the 6-man tables, 3GB (the practical limit of 32-bit machines) is insufficient.
Outline - is conservative by intention.
step 0. Complete work on tbstats such that it can read compressed files directly.
step 1. Consistent extension of indexing scheme in tbgen.
step 2. Reduction of memory requirements via block processing.
step 3. Speed improvements; perhaps by adopting the work of Konoval, Bourzutschky, and Muller.
step 4. Extensions to support complete 7-man tables.
john
Posted: Sat Jun 23, 2007 4:09 pm
by Kirill Kryukov
Hi John and everyone, thanks for looking into this problem. I can contribute CPU time of a machine with 8 GB RAM, if it can be useful. (Windows XP Professional 64-bit).
Posted: Sat Jun 23, 2007 9:55 pm
by ZeroOne
jkominek wrote:step 4. Extensions to support complete 7-man tables.
That will get short one day too. Why not program it to do
n-men tables?
As a side note, my computer has 1.5 GB of RAM.
Posted: Wed Jul 04, 2007 5:17 am
by vb4
jkominek wrote:Hi Ernest,
Way back in early January, still feeling the liberation of Christmas holidays, I made headway in the task of extending tbgen for this purpose. In a spate of New Years resolutions of
Great Progress This Year I had, well, uh, visages of great progress -- of the generation code churning away for months on end, of emule-ing files for others to verify, of generating 5-1 stats that "vb4" would so much like to have hands on. Then came February. Poof.
Hi John,
I am already keeping my fingers crossed! I hope that someone will get the ball rolling for this 5-1 egtb set.
Thanks for remembering my interest,
Les
(My "all-day and all-night job," is being a computer science PhD student at a prestigious university. "Prestigious" means you get worked like a dog for someone else's glory, for an interminable amount of time. Until your funding runs dry. Then you get kicked out. Details at
http://www.phdcomics.com)
I see some relief in sight, round about early July. So, events willing, expect another posting then. Maybe I'll have something concrete.
In the interim, I'm open to pledges of donated computing power. Sucking energy out of our compute cluster as school would not be kosher, and my personal machines are showing wrinkles around the corners of their eyes (just like me). I haven't been on this board much in the past several months, but, having announced my intention here, will check back for any takers.
john
Re: 5-1 EGTB Status. What is needed to make them useable?
Posted: Thu Aug 16, 2007 4:38 am
by jkominek
Hi everyone,
What has it been? Nearly two months since last making a squawk on these boards. Oh my. New server, new look! So much has happened since. And where have I been exactly, after making a pledge to be a more active member?
Turns out the project I expected to end in June pushed well into July. At which point I was told it would be a Good Idea to propose. (My thesis; not the other kind.) Around these parts that turns out to be a Big Deal, with no small amount of panic too. And it's happening ... tomorrow. Erk -- make that today, now that it is past midnight from where I write.
Happy to see there is much to catch up on once I can come up for breath.
john
Re: 5-1 EGTB Status. What is needed to make them useable?
Posted: Sun Aug 26, 2007 6:23 pm
by jshriver
I'd also be willing to donate CPU time. If you can send me a tarball or zip of your modified tbgen I'll get cracking
-Josh
N-man table support
Posted: Wed Mar 12, 2008 7:14 am
by jkominek
ZeroOne wrote:jkominek wrote:step 4. Extensions to support complete 7-man tables.
That will get short one day too. Why not program it to do
n-men tables?
I suppose it is because I don't know how to do that. At least not easily. The only solution I know of is meta-programming. The writing of one program (tbgen) with another program (tbgengen). A single generic program that can handle N-men would almost surely concede unacceptable space and time efficiency.
john
Re: 5-1 EGTB Status. What is needed to make them useable?
Posted: Wed Mar 12, 2008 1:07 pm
by Codeman
I am working on a very general bitbase generator that works with n-men. But still there are some problems for >6 men egtbs. Till now all Tablebases could be generated mainly from ram for larger tablebases new solutions will have to be found.
Re: 5-1 EGTB Status. What is needed to make them useable?
Posted: Wed Mar 12, 2008 4:33 pm
by jkominek
Codeman wrote:I am working on a very general bitbase generator that works with n-men.
Since you are thinking very general, can you make the position generator work for arbitrary WxH rectangular chess boards, with and without castling and en passant?
john
Re: 5-1 EGTB Status. What is needed to make them useable?
Posted: Wed Mar 12, 2008 6:31 pm
by Codeman
jkominek wrote:Since you are thinking very general, can you make the position generator work for arbitrary WxH rectangular chess boards, with and without castling and en passant?
john
I am a fan of bitboards so the whole code is based on them. But the algorithm in principle would as well meet your requirements.