End Game Tablebase Files (5-Man) For Crafty
Robert Hyatt
: Can someone give me an estimate of the total amount of disk
: space consumed by the five man end game files for Crafty? Downloading
: them is going be a considerable investment of time, and would be
: useless unless I have the space to decompress them afterwards.
there are currently about 5.5 gigs (compressed) available. I believe
that Eugene said that the 3-4 piece files, plus kxxkx, are going to
take about 20 gigs uncompressed. He is eventually going to offer
on-the-fly uncompression as well, but that will not obsolete your
downloads, because he is going to take the standard (uncompressed)
files and compress those using a separate utility...
--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
Robert Hyatt
: Doy Davis <d...@mindspring.com> wrote:
: : Can someone give me an estimate of the total amount of disk
: : space consumed by the five man end game files for Crafty? Downloading
: : them is going be a considerable investment of time, and would be
: : useless unless I have the space to decompress them afterwards.
: there are currently about 5.5 gigs (compressed) available. I believe
: that Eugene said that the 3-4 piece files, plus kxxkx, are going to
: take about 20 gigs uncompressed. He is eventually going to offer
: on-the-fly uncompression as well, but that will not obsolete your
: downloads, because he is going to take the standard (uncompressed)
: files and compress those using a separate utility...
Damn... did I write that???
make that on-the-fly *decompression*...
:)
Komputer Korner
Thompsons go on 4 CDs so his can't be any greater than 2.5Gb
compressed. Of course not every 5 man is on those 4CDs.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Robert Hyatt wrote in message <7315dd$bbc$2...@juniper.cis.uab.edu>...
>Doy Davis <d...@mindspring.com> wrote:
>: Can someone give me an estimate of the total amount of disk
>: space consumed by the five man end game files for Crafty?
Downloading
>: them is going be a considerable investment of time, and would be
>: useless unless I have the space to decompress them afterwards.
>
>there are currently about 5.5 gigs (compressed) available. I believe
>that Eugene said that the 3-4 piece files, plus kxxkx, are going to
>take about 20 gigs uncompressed. He is eventually going to offer
>on-the-fly uncompression as well, but that will not obsolete your
>downloads, because he is going to take the standard (uncompressed)
>files and compress those using a separate utility...
>
>
>
Eugene Nalimov
If you have not enough disk space, you can sort them
depending of their usefulness in the actual game, and
download only top N.
Also please note that if you'll use KXPKY, you'll need
4 K*PKY; size of entire set will be ~1.1-1.2Gb. So,
mentioned above N usually will be in 0-5 range.
Eugene
Doy Davis wrote in message <3653c1a0...@news.mindspring.com>...
Eugene Nalimov
Eugene Nalimov wrote in message <7322cf$8...@news.dns.microsoft.com>...
>22.5Gb uncompressed when all 3+2 TBs will be available.
>If you have not enough disk space, you can sort them
>depending of their usefulness in the actual game, and
>download only top N.
>
>Also please note that if you'll use KXPKY, you'll need
>4 K*PKY; size of entire set will be ~1.1-1.2Gb. So,
Sorry, of course I meant 4 KX*KY, where * in (N,B,R,Q).
Eugene
Robert Hyatt
: 22.5Gb uncompressed when all 3+2 TBs will be available.
: If you have not enough disk space, you can sort them
: depending of their usefulness in the actual game, and
: download only top N.
: Also please note that if you'll use KXPKY, you'll need
: 4 K*PKY; size of entire set will be ~1.1-1.2Gb. So,
: mentioned above N usually will be in 0-5 range.
: Eugene
Eugene...
a couple have claimed we already have over 30 gigs on my ftp when they
are uncompressed. Is something odd going on???
Anders Thulin
Komputer Korner <kor...@netcom.ca> wrote:
>What will be the compressed total for all 3,4,5 man endgames? Ken
>Thompsons go on 4 CDs so his can't be any greater than 2.5Gb
>compressed. Of course not every 5 man is on those 4CDs.
There are only 3 Thompson CD's, as far as I know. They contain no
4+1 endgames, and some endgames appear on 2 CD's -- in one or two
cases, on all three (KNN-KN is one of those, I think).
Note, though, that there are only black-to-move files. Each 5-piece
file also contains some (all?) of the 4- and 3- piece subendgames
(except promotions), so it's not trivial to make direct comparisons.
Based on earlier experiments with compressing the Thompson and
Stevens endgame files, I'd guess that the size of the compressed files
will be approximately the same as the sum total of the gzipped files,
or slightly less. Thus, approx 5 Gb storage will be probably required
for all files.
However, the Thompson and Stevens files were easier to compress --
all files had internal structures that were multiples of 64 bytes. One
single compression method was enough to do the job for all files --
although of course the code tables for each file would be different.
In some cases, several multi-kbyte chunks of the files were all
illegal positions, and so could be compressed to essentially one
single byte each.
The Nalimov files have much more complex internal structure, with
far less such obvious redundance, and so probably will require
specialized compression methods for each class of endgames.
Unless a general compression method like gzip is used, of course.
--
Anders Thulin Anders....@telia.se 013-23 55 32
Telia ProSoft AB, Teknikringen 6, S-583 30 Linkoping, Sweden
Ernst A. Heinz
> There are only 3 Thompson CD's, as far as I know.
There are 4 Thompson CDs (see the LaTeX source at the end of this
message which constitutes an appendix of my article "Endgame Databases
and Efficient Index Schemes for Chess" as submitted to the ICCA Journal).
> They contain no 4+1 endgames,
Wrong, KNNNK is included on CD #2.
> Note, though, that there are only black-to-move files. Each 5-piece
> file also contains some (all?) of the 4- and 3- piece subendgames
> (except promotions), so it's not trivial to make direct comparisons.
They contain *all* subgames except promotions. You probe the subgames by
virtually placing the pieces to be removed from the board on the same
square as the white King.
=Ernst=
+----------------------------------------------------------------------------+
| Ernst A. Heinz, School of CS (IPD), Univ. of Karlsruhe, P.O. Box 6980, |
| D-76128 Karlsruhe, F.R. Germany. WWW: <http://wwwipd.ira.uka.de/~heinze/> |
| Tel. / Fax: +49-(0)721-6084386 / 6087343 E-Mail: <hei...@ira.uka.de> |
+----------------------------------------------------------------------------+
"It has recently been found out that research causes cancer in rats!"
//////////////////////////////////////////////////////////////////////////////
%***********************************************
\section{Thompson's Endgame Databases on CD-ROM}
%***********************************************
\noindent
Until he finally ran out of supply in 1996, Thompson himself distributed
his endgame databases on CDs to anybody who expressed interest in them.
Right now, the cheapest and easiest way to aquire Thompson's complete
set of endgame databases seems to buy Meyer-Kahlen's commercial chess
program {\sc Mephisto Shredder} that includes Thompson's 4 CDs and sells
for 99~DM $\approx$ 55 US-\$ in Germany. The whole {\sc Shredder} family
and their commercial competitors {\sc Fritz~4\,}~\& {\sc Fritz~5\,} by
Morsch belong to the small group of strong chess programs that still use
Thompson's databases. Of course, they probe the compressed databases
only at the root position.
Below follows a complete list and a cross-reference of all 5-piece
endgame databases as constructed by Thompson and available on each CD.
The given database identifiers name the pieces of the potentially
winning side followed by an ``\_'' character and the pieces of the
opponent while omitting both Kings (e.g.\ BB\_N denotes KBBKN). The
numbers in brackets of the cross-reference tell you which CDs contain
the according database. Some are actually present multiple times in
order to make each CD self-contained. Please also note that Thompson's
5-piece databases include all of their 3-piece and 4-piece sub-games
except for those resulting from promotions of Pawns. By assigning the
location of the potentially winning King to all pieces that are no
longer present on the board, you can easily probe all sub-games of any
database in a convenient way without switching files.
\begin{description}
\item[{\bf Thompson CD \#1:\ }]
BB\_N, BN\_N, NN\_N, NP\_N, Q\_BB, Q\_BN, QB\_Q, QN\_N, Q\_NN, QN\_Q,
QP\_Q, QQ\_Q, Q\_RB, Q\_RN, QR\_Q, QR\_R, Q\_RR, RB\_R, RN\_N, RN\_R,
RP\_R, RR\_R.
\item[{\bf Thompson CD \#2:\ }]
BB\_B, BB\_N, BN\_B, BN\_N, BP\_B, BP\_N, NN\_B, NNN\_,\,NP\_B, QB\_B,
QB\_N,\,QN\_B, QR\_Q, RB\_B, RB\_N, RB\_Q, RN\_B, RN\_Q, RP\_Q, RR\_Q.
\item[{\bf Thompson CD \#3:\ }]
NN\_B, NN\_N, NN\_P, NN\_Q, NN\_R, Q\_QR, QR\_B, Q\_RB, Q\_RN, Q\_RP,
Q\_RR, RB\_B, R\_BB, R\_BN, R\_BP, RN\_B, R\_NN, R\_NP, RP\_B, R\_QB,
R\_QN, RR\_B, R\_RB, R\_RN.
\item[{\bf Thompson CD \#4:\ }]
BN\_B, BN\_N, BN\_P, BN\_Q, BN\_R, Q\_BB, Q\_BN, Q\_BP, Q\_QB,
Q\_RB, QR\_N, RB\_N, RN\_N, RP\_N, RR\_N.
\item[{\bf Cross-Reference:\ }]
BB\_B[2], BB\_N[1:2], BN\_B[2:4], BN\_N[1:2:4], BN\_P[4], BN\_Q[4],
BN\_R[4], BP\_B[2], BP\_N[2], NN\_B[2:3], NNN\_[2], NN\_N[1:3],
NN\_P[3], NN\_Q[3], NN\_R[3], NP\_B[2], NP\_N[1], QB\_B[2], Q\_BB[1:4],
QB\_N[2], Q\_BN[1:4], Q\_BP[4], QB\_Q[1], QN\_B[2], QN\_N[1], Q\_NN[1],
QN\_Q[1],\,QP\_Q[1],\,Q\_QB[4],\,QQ\_Q[1],\,Q\_QR[3],\,QR\_B[3],
Q\_RB[1:3:4], QR\_N[4], Q\_RN[1:3:4], Q\_RP[3], QR\_Q[1:2], QR\_R[1],
Q\_RR[1:3], \\ RB\_B[2:3], R\_BB[3], RB\_N[2:4], R\_BN[3], R\_BP[3],
RB\_Q[2], RB\_R[1], RN\_B[2:3], RN\_N[1:4], R\_NN[3], R\_NP[3],
RN\_Q[2], RN\_R[1], RP\_B[3], RP\_N[4], RP\_Q[2], RP\_R[1], R\_QB[3],
R\_QN[3], RR\_B[3], R\_RB[3], R\_RN[3], RR\_N[4], RR\_Q[2], RR\_R[1].
\end{description}
Eugene Nalimov
>Eugene Nalimov <eug...@microsoft.com> wrote:
>: 22.5Gb uncompressed when all 3+2 TBs will be available.
>: If you have not enough disk space, you can sort them
>: depending of their usefulness in the actual game, and
>: download only top N.
>
>: Also please note that if you'll use KXPKY, you'll need
>: 4 K*PKY; size of entire set will be ~1.1-1.2Gb. So,
>: mentioned above N usually will be in 0-5 range.
>
>: Eugene
>
>Eugene...
>
>a couple have claimed we already have over 30 gigs on my ftp when they
>are uncompressed. Is something odd going on???
Something must be totally wrong... I calculated size for each possibleTB
(program use them for simple sanity check during startup), and total size
is 22.5Gb w/o 4+1 tables, 29.9 with them.
Actual sizes for each generated table is exactly that pre-calculated one,
so I don't think the error is mine.
Eugene
Robert Hyatt
: Robert Hyatt wrote in message <732lg4$obg$2...@juniper.cis.uab.edu>...
:>Eugene Nalimov <eug...@microsoft.com> wrote:
:>: 22.5Gb uncompressed when all 3+2 TBs will be available.
:>: If you have not enough disk space, you can sort them
:>: depending of their usefulness in the actual game, and
:>: download only top N.
:>
:>: Also please note that if you'll use KXPKY, you'll need
:>: 4 K*PKY; size of entire set will be ~1.1-1.2Gb. So,
:>: mentioned above N usually will be in 0-5 range.
:>
:>: Eugene
:>
:>Eugene...
:>
:>a couple have claimed we already have over 30 gigs on my ftp when they
:>are uncompressed. Is something odd going on???
: Something must be totally wrong... I calculated size for each possibleTB
: (program use them for simple sanity check during startup), and total size
: is 22.5Gb w/o 4+1 tables, 29.9 with them.
: Actual sizes for each generated table is exactly that pre-calculated one,
: so I don't think the error is mine.
: Eugene
turned out to be operator error. :) With that many digits, he was
apparently seeing 30B rather than 3B. :)
Bob
:>
:>--
Anders Thulin
Ernst A. Heinz <hei...@ira.uka.de> wrote:
>> There are only 3 Thompson CD's, as far as I know.
>There are 4 Thompson CDs
Apparently. When did the 4th come? I bought mine from Ken Thompson
in March 1994; at that time there were only 3, and no more volume was
planned as `the games left out are not important enough to make
another.'
>They contain *all* subgames except promotions.
So it makes it even more difficult to compare the Nalimov files and
the Thompson files by size.