Author: Guido
Date: 13:32:54 04/02/02
Go up one level in this thread
On April 02, 2002 at 13:27:10, Mike Hood wrote:
>The normal figure thrown about for the size of a complete 6-piece egtb
>collection is "about a terabyte". Can we try to make this figure a bit more
>accurate? In the tablebase generation program the file size calculation is
>separate from the generation itself, so it should be possible to calculate the
>size of the uncompressed files exactly. When we have that we can make an
>estimation of the size of the compressed tablebases (approx 20% of the
>uncompressed size). And when we have that figure we can sit back and watch the
>developments in storage media to see when 6-piece TBs become technically
>feasible. Constellation 3D promised their FMD technology would produce optical
>disks with a capacity of 1.5 terabyte, but they've failed to deliver so far.
>
>There should be 365 6-piece tablebase pairs of varying importance, in three main
>groups:
>
>Kxx-Kxx : 120 tablebases
>Kxxx-Kx : 175 tablebases
>Kxxxx-K : 70 tablebases
>
>If we have the size for each group we'll know how much space a person needs on
>his hard drive to generate KPPKPP, KPPPKP and KPPPPK respectively. Or someone
>can fire up his university's supercomputer.
>
>Maybe it's just an academic pursuit at present, but technology has a habit of
>overtaking you when you least expect it.
>
>And when we've finished with the 6-piece tablebases, what about the 1001 7-piece
>egtb's?
>
>Kxxx-Kxx : 525 tablebases
>Kxxxx-Kx : 350 tablebases
>Kxxxxx-K : 126 tablebases
>
>I bet you're all dying to get your hands on the KBBBBBK tablebase ;)
For uncompressed endings I obtain these values:
No. men Different Space
Endings in Mb (1Mb = 1024^2)
2 1 ---
3 5 0.275 (kbk and knk excluded)
4 30 152.333
5 110 31,865.260
6 365 5,150,003.501
7 1001 611,686,390.198
8 2520 59,916,917,072.203
Values reported are the result of an exact calculation using double precision
floating point, but they_are_depending_on_the_indexing_scheme adopted.
So they can be larger or smaller of the correspondent values obtained by other
indexing scheme.
Of course if the program is correct ...
In details for 6 and 7 men:
Kxx-Kxx 1,992,702.965 120
Kxxx-Kx 2,525,840.429 175
Kxxxx-K 631,460.107 70
------------- ---
5,150,003.501 365
Kxxx-Kxx 382,303,993.874 525
Kxxxx-Kx 191,151,996.937 350
Kxxxxx-K 38,230,399.387 126
--------------- ----
611,686,390.198 1001
Ciao
Guido
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.