Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: which 6 man tablebases are the most important?

Author: Vincent Diepeveen

Date: 12:47:57 04/02/04

Go up one level in this thread


On April 02, 2004 at 15:24:27, Jonas Bylund wrote:

>>DIEP is having its own GUI nowadays (apart from that the engine is in
>>cardozachess too) and has its own EGTB's which are a lot smaller.
>
>A couple of questions:
>
>1)Will Diep ever be available for windows (GUI and all) personally i would pay
>good money for that.

Thanks for your kind questions. Not exactly like Uri Blass who just want to
download free engines :)

Diep just runs under windows indeed.

If i understand well microsoft has copyrighted their DeviceContexts. That means
that other manufacturers like SGI are not allowed to release libraries that can
cast onto the DC so that things like compatible OpenGL drivers will work in
windows.

So any Open GL drivers after 1.1 are illegal in that sense.

Most people do not realize what this means for the future of mass software
products. IMHO it can only mean that you must release mass software market
products for windows. The rest doesn't make a chance.

Additionally shops do not want to sell linux games at all. I know why, i use
daily linux myself so i am qualified to answer. In fact i started using linux in
1993/1994 if i remember well. It was version 1.0, still have the cdrom here.

It was the first days of linux. It has improved a lot, but it is still too
difficult to maintain for a normal user. Especially hardware support is very
poor for the latest hardware.

I need to recompile the kernel to just get my $10 network cards to work and i'm
not sure whether my sound cards work anyway. Any professional cards in whatever
sense forget it, won't work at all.

So Linux where it is moving fast and i hope will get professional, for the mass
software market it has completely lost already.

We are of course working hard at DIEP to release it somewhere in 2004.

>2)What is the size of your format when combining all 3+4+5+6 men tb's?

I'm still busy in the process of getting them smaller.

The compression is very important which you use.

Look there is some superior compression around on this planet. Like RKC.

That compressing eats a lot of cpu time is not so interesting, but decompressing
takes really a lot of time.

Kadatch is doing a really great job there.

I have some ideas to write some compression myself, but in advance i must admit
that odds are like 0.1% that i will outperform Kadatch.

I'm having some compression experiences already. At university i experimented a
lot with compression, so i'm not the complete layman like some might think.

But i'm of course a complete layman when compared to Kadatch.

If i see what he manages to achieve with just Huffman then that is real
impressive.

Only after those experiments i will be able to give a real answer.

For now assume i will have completely failed if it is not far under 10GB.

Factor 70 smaller than Nalimov is doable though.

The real pain in any prediction on how big they will get, is the 33p.

Look everyone understands that certain egtb's compress real well:

28-03-2004  01:12           546.085 kqbbkp_w.dtb.emd
28-03-2004  01:32           534.702 kqnnkp_w.dtb.emd
28-03-2004  02:15           218.108 kqqqkp_w.dtb.emd
20-03-2004  18:47           203.028 kbbbkp_b.dtb.emd
24-03-2004  02:54           195.274 knnnkp_b.dtb.emd
28-03-2004  01:35           153.335 krrrkp_w.dtb.emd
28-03-2004  02:14           134.072 kqqqkp_b.dtb.emd
28-03-2004  01:34           130.540 krrrkp_b.dtb.emd

Only huge changing dictionaries can beat the above sizes.

Uncompressed in diep format most of them are 600MB.

But this is 42p.

Everyone understands that qqq versus pawn is an easy win.


However

KRP KRP

Let's be clear. How well will that compress?

I do not know.

It might be a big big problem to compress.

I foresee more disasters with 33p. Let's wait and see.

In that sense the 7 men will be easier.

>3)When you say you have tb's that does not take as much space as the publicly
>available, why not make it publicly available with source code, so programmers
>can start to implement support?

Now here i finally will give an answer understood by Uri Blass very well i
guess:

What is the advantage for me to do that?

>Regards
>Jonas



This page took 0.01 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.