Glaurung 2-epsilon

[Tord Romstad]

For those brave enough to test an unfinished and possibly unstable new program, Glaurung 2-epsilon ("two minus epsilon") is now available for download from the Glaurung web site. At the moment, only the source code and binaries for Mac OS X are included. Windows and Linux executable will follow soon.

This is a complete rewrite of Glaurung; the name is almost the only thing that remains of the old program. The new program is still far from complete, and lacks important features like parallel search, multi-PV mode, and configurable engine parameters. The style of play is dull, passive and ugly because of lack of chess knowledge. Don't expect to see any attractive attacking games in the style of Glaurung 1.x.

If this doesn't put you off, give it a try. I think the current version is at least as strong as Glaurung 1.2.1 (assuming that both engines use only a single CPU). The missing features and knowledge will be added gradually, beginning with the parallel search.

Have fun!

Tord

[Graham Banks]

Hi Tord,

do you think it will be a long wait for us before the finished product?

Regards, Graham.

[Yar]

good news !

[Tord Romstad]

Hi Tord,

do you think it will be a long wait for us before the finished product?

Hi Graham,

That's a very difficult question. My life is notoriously unpredictable, and my ambitions and priorities change all the time. As a rough estimate, I guess you will have to wait at least six weeks and at most six years, but I wouldn't bet on it.

Glaurung 2-epsilon/2, which will add parallel search, is probably not very far off, though. It will probably be ready some time this summer; precisely how soon depends on the frequency of rainy weekends.

For serious tournaments on single-CPU computers, I currently recommend using Glaurung 2-epsilon rather than Glaurung 1.2.1. As soon as 2-epsilon/2 is ready, I recommend using that version in all tournaments. For entertainment and analysis use, I recommend to keep using 1.2.1 a bit longer.

It is probably best to wait a bit longer before testing the new Glaurung at slow time controls on the CCRL list. I would be happy to see it tested on your blitz list, but I understand and respect if you have other priorities.

Tord

[Vempele]

Now compiled for Windows and playing here. Thanks!

[Tord Romstad]

Now compiled for Windows and playing here.

Great!

What compiler did you use, and what changes (in any) did you have to do in order to make the program compile cleanly?

Tord

[Vempele]

Now compiled for Windows and playing here.

Great!

What compiler did you use, and what changes (in any) did you have to do in order to make the program compile cleanly?

Tord

Microsoft Visual Studio Codename "Orcas" beta 1 - no changes made. It gave 179 warnings, though. Could you post what the analysis for the starting position is supposed to look like after a few plies? Here's my result:

2 00:00 46 1.642 +0,11 Nf3 Nf6
3 00:00 185 6.379 +0,54 Nf3 Nf6 Nc3
4 00:00 378 13.034 +0,11 Nf3 Nf6 Nc3 Nc6
5 00:00 796 26.533 +0,15 Nf3 Nf6 Nc3 Nc6 d3
6 00:00 1.633 51.031 +0,11 Nf3 Nf6 Nc3 Nc6 d3 d6
7 00:00 4.213 120.371 +0,25 Nf3 Nf6 Nc3 Nc6 e3 d6 Bd3
8 00:00 7.396 189.641 +0,11 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5
9 00:00 20.983 287.438 +0,45 Nf3 Nf6 Nc3 Nc6 e3 Nb4 a3 Nbd5 Bd3
10 00:00 43.081 410.295 +0,11 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5 0-0 0-0
11 00:00 117.374 569.776 +0,43 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bd6 0-0 0-0 Ng5
12 00:00 210.892 644.929 +0,35 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5 0-0 0-0 Ng5 d5
13 00:00 529.773 695.240 +0,21 Nf3 Nf6 Nc3 Nc6 d4 e6 e4 Bb4 Bd3 0-0 0-0 d5 exd5 exd5
14 00:01 1.207.979 652.608 +0,25 Nf3 Nf6 Nc3 Nc6 d4 e6 e4 Bb4 Bd3 0-0 0-0 d5 exd5 exd5 Bf4 Bxc3 bxc3
15 00:05 3.475.555 648.545 +0,15 Nf3 Nf6 Nc3 d5 e3 e6 d4 Nc6 Bd3 Be7 0-0 0-0 Bd2 Nb4 Bb5

[Tord Romstad]

Microsoft Visual Studio Codename "Orcas" beta 1 - no changes made. It gave 179 warnings, though.

Thanks for the info! 179 warnings is a lot. Clearly the Microsoft compiler and GCC give very different warnings. I compile with the "-Wall" switch (all warnings enabled) with GCC, and there are no warnings at all...

Could you post what the analysis for the starting position is supposed to look like after a few plies? Here's my result:

2 00:00 46 1.642 +0,11 Nf3 Nf6
3 00:00 185 6.379 +0,54 Nf3 Nf6 Nc3
4 00:00 378 13.034 +0,11 Nf3 Nf6 Nc3 Nc6
5 00:00 796 26.533 +0,15 Nf3 Nf6 Nc3 Nc6 d3
6 00:00 1.633 51.031 +0,11 Nf3 Nf6 Nc3 Nc6 d3 d6
7 00:00 4.213 120.371 +0,25 Nf3 Nf6 Nc3 Nc6 e3 d6 Bd3
8 00:00 7.396 189.641 +0,11 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5
9 00:00 20.983 287.438 +0,45 Nf3 Nf6 Nc3 Nc6 e3 Nb4 a3 Nbd5 Bd3
10 00:00 43.081 410.295 +0,11 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5 0-0 0-0
11 00:00 117.374 569.776 +0,43 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bd6 0-0 0-0 Ng5
12 00:00 210.892 644.929 +0,35 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5 0-0 0-0 Ng5 d5
13 00:00 529.773 695.240 +0,21 Nf3 Nf6 Nc3 Nc6 d4 e6 e4 Bb4 Bd3 0-0 0-0 d5 exd5 exd5
14 00:01 1.207.979 652.608 +0,25 Nf3 Nf6 Nc3 Nc6 d4 e6 e4 Bb4 Bd3 0-0 0-0 d5 exd5 exd5 Bf4 Bxc3 bxc3
15 00:05 3.475.555 648.545 +0,15 Nf3 Nf6 Nc3 d5 e3 e6 d4 Nc6 Bd3 Be7 0-0 0-0 Bd2 Nb4 Bb5

I get the same node counts and PVs for the first 6 iterations, after that they diverge. Perhaps the difference is caused by different hash table sizes (I used the default setting of 32 MB). Here is my output:

Code: Select all

2 +11 0 46 Nf3 Nf6
3 +54 0 185 Nf3 Nf6 Nc3
4 +11 0 378 Nf3 Nf6 Nc3 Nc6
5 +15 0 796 Nf3 Nf6 Nc3 Nc6 d3
6 +11 1 1633 Nf3 Nf6 Nc3 Nc6 d3 d6
7 +25 1 4102 Nf3 Nf6 Nc3 Nc6 e3 d6 Bd3
8 +11 2 7288 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5
9 +56 5 24490 Nf3 Nf6 Nc3 Nc6 e3 Nb4 Ne5 Nc6 Nxc6 bxc6
10 +11 12 59106 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bc5 O-O O-O
11 +43 27 131921 Nf3 Nf6 Nc3 Nc6 e3 e6 Bd3 Bd6 O-O O-O Ng5
12 +35 61 309410 Nf3 Nf6 e3 e6 Bd3 Nc6 O-O Bc5 Nc3 O-O Ng5 d5
13 +17 119 603067 Nf3 Nf6 e3 e6 Bd3 Nc6 O-O Bc5 Nc3 O-O Na4 Bd6 Ng5
13 +21 232 1158986 e4 Nf6 Nc3 Nc6 Nf3 e6 d4 Bb4 Bd3 O-O O-O d5 exd5 exd5
14 +21 598 2898273 e4 Nf6 Nc3 Nc6 Nf3 e6 d4 Bb4 Bd3 O-O O-O Bxc3 bxc3 d5 Ba3
15 +25 1029 5021775 e4 Nf6 Nc3 Nc6 Nf3 e6 d4 Bb4 Bd3 O-O O-O d5 exd5 exd5 Bf4 Bxc3 bxc3
16 +35 3469 16605733 e4 Nf6 e5 Nd5 Nf3 e6 Bc4 Nb6 Be2 Nc6 O-O Be7 Nc3 O-O d4 d5

Tord

[Vempele]

In console mode: depths 1-13 identical to my first run, then

14. 1207972
15. 3475222
16. 14908970

[Shaun]

Hi Tord,

This is great news - once there is an official windows exe I will start testing

All the best and lookign for a wet summer

Shaun