Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To Robert Hyatt a tablebase space

Author: Eugene Nalimov

Date: 20:35:54 11/07/01

Go up one level in this thread


Problem with compression is simple: TB is compressed in 8Kb blocks, and each
block is compressed "from the scratch", so you can decompress it individually.
Decompression code allows larger block sizes. When block size is 64Kb,
compression is much better, significally better than zip/gzip/etc [because
Andrew Kadatch, who wrote TB compression/decompression code, used algorithm that
are well-suited for the TBs].

The problem is that 64Kb is too much. Each TB probe that misses TB cache becomes
~8 times slower (actually more like 5 or 6, as, due to better compression, you
read from disk not 8, but "only" 5 or 6 times more).

Bob run a lot of experiments before we came to 8Kb blocks. That is good
compromise between TB probing speed and compression ratio.

Eugene

On November 07, 2001 at 18:59:12, Thomas Mayer wrote:

>Hi Jonas,
>
>>>Hi Robert, i need your help about the tablebase in your ftp. I download the
>>>folder FIVE and this have 7GB!! of information (a total of 10 CD of 700 Mb).
>>>Now the 9 CD of Nalimov tablebase contain the 3, 4 and 5 pieces final.
>
>there is a difference, some Nalimov collections are without the 4+1 stones, some
>or with them - afaik Bobs collection on the server has the 4+1 stones. Usually
>you need 12 CDs for all 3,4+5 stones. Other collection are somehow incomplete.
>
>>Any posiblity of making TB's readable from .zip files?
>>(maby a stupid quetion, but it would cut the sizes in half)
>
>The Nalimovs are already compressed... but for fun I have tested it, it get's
>nearly a half percent out of them... impressive... :)
>
>Greets, Thomas



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.