Glaurung 2-epsilon/5

[Tord Romstad]

Hello folks,

As some of you may have noticed already, Glaurung 2 - ε/5 has been available from my site since yesterday night. This isn't a very exciting update, unless you happen to be using an Intel Macintosh. The Macintosh executable is now compiled with the Intel compiler, and is a lot faster than the GCC compiled executable of the previous version.

The most important change in the code is that I have tried to fix a bug reported by Werner Schüle: Glaurung lost several CEGT games on time just before giving mate (which is of course an unfortunate moment to lose on time). Another change worth mentioning is that the parallel search has been made slightly more efficient; on my computer the speed has improved by about 10% with two search threads. I have some hope that the improvement will be bigger on systems with 4 CPUs, but I haven't had the chance to verify this.

For the other (very minor) changes, have a look at the Readme.txt file.

For those of you who produce rating lists: It probably isn't worth the effort to start your tests from scratch with Glaurung 2 - ε/5. It's basically the same as version 2 - ε/4, with just a few little bug fixes and microscopic enhancements. I am almost 100% sure the new version is no stronger than the old (unless you use a Macintosh with an Intel CPU, of course). If your rules allow it, please just silently replace 2 - ε/4 with 2 - ε/5, and let the new version inherit the rating of the old.

This will probably be the last version for a while, unless there are some very ugly bugs. From now on, I'll concentrate on writing a half decent evaluation function (the current one is terrible), add back most of the features from Glaurung 1.x, and try to complete my GUI. I expect the final 2.0 version to be ready some time during the autumn.

Tord

[ArmyBridge]

Hi Tord, I wish you good luck with your next version, congratulation because your engine is very good,I just Hope that you would teach to next version to fish a small fish you know I am a bit tired of Rybka squash all the competition... Regards

[Tord Romstad]

Hi Tord, I wish you good luck with your next version, congratulation because your engine is very good,I just Hope that you would teach to next version to fish a small fish you know I am a bit tired of Rybka squash all the competition... Regards

Hello Armando,

Thanks for your kind words! I doubt that the final Glaurung 2 will score many points against Rybka or other top programs, but at least I expect it to play better than the current development version, and to have a more attractive style of play.

Tord

[Tord Romstad]

Louis Zulli has made me aware of an Intel compiler switch I didn't know about (-mdynamic-no-pic) which enabled me to increase the speed of the Mac OS X binary by about 8%!

Intel Mac users who want the best possible performance are recommended to download Glaurung 2 - ε/5 again.

Tord

[ArmyBridge]

Tord, maybe your engine wolud not win a Match vs Rybka, but for me if I would see Glaurung beat Rybka in some game in great style is enough!! the most important thing is to enjoy the beauty of the chess and Glaurung has the touch . It has made great & beauty games

[glorfindel]

I have very high expectations for Glaurung 2.0. I think it will be one of the very best.

Good luck and good inspiration.

[BBauer]

Hi everybody,

in the following game

[White "J. Hartweg"]
[Black "N.N."]
[Result "1-0"]
[PlyCount "51"]

1. e4 e5 2. Nc3 Nc6 3. f4 exf4 4. Nf3 g5 5. h4 g4 6. Ng5 h6
7. Nxf7 Kxf7 8. d4 d5 9. Bxf4 dxe4 10. Bc4 Kg6 11. O-O Bg7
12. h5 Kh7 13. Be5 Bxe5 14. Rf7 Bg7 15. Nxe4 Qxd4
16. Qxd4 Nxd4 17. Raf1 Bf5 18. c3 b5 19. cxd4 bxc4
20. R1xf5 Re8 21. Re5 Rf8 22. Ree7 Rxf7 23. Rxf7 1-0

black (a human) resigned after 23. Rxf7.

[D] 6nr/p1p2Rbk/7p/7P/2pPN1p1/8/PP4P1/6K1 b - - 7 23

On an AMD 64 X2 Dual Core Processor 5200+, 2.6GHz it took
glaurung2-epsilon/5 8:19 to see it is mated.

Code: Select all

  20	00:12	  30.912.639	2.523.480	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  21	00:18	  49.162.816	2.593.933	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  22	00:35	  93.432.205	2.639.999	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  23	01:04	 172.684.871	2.668.225	-1,31	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  24	03:04	 491.877.670	2.670.983	-1,45	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  25	08:19	1.334.990.761	2.671.680	-M6	c7c6 Kg1f2 a7a5 Kf2g3 a5a4 a2a3 c4c3 b2xc3 c6c5 d4xc5 Ng8f6 Ne4xf6+

How long does other programs need to see it's over?

By the way, how can I start a new thread in this forum?

kind regards
Bernhard

[Sylwy]

"By the way, how can I start a new thread in this forum? "

Push "NewTopic" button.

[BBauer]

Thank you very much!

[Stan Arts]

It's a zugzwang position.
Neurosis 2.3 on PIV 3.4GHz 128MB hash. :

Code: Select all

Legal moves: 8   Static score: 236
Ply Score  Time     Nodes      Best move and expected line
------------------------------------------------------------
10  37     78       152083     a6 Rxc7 g3 Kf1 a5 Nxg3 Nf6 Rxc4 Rd8 Ke2 Re8+ Kd3
11  13     89       181198     a6 < - > Rxc7 g3 Kf1 a5 a4 Ne7 Rxe7 Rd8
11  13     109      232479     a6 Rxc7 g3 Kf1 a5 a4 Ne7 Rxe7 Rd8
11  14     153      372351     c6 < + > Rxa7 g3 a4 Ne7 Rxe7 Kg8 Bxd4+ Kf1
11  14     153      372540     c6 Rxa7 g3 a4 Ne7 Rxe7 Kg8 Bxd4+ Kf1
12  -10    206      516910     c6 < - > Rxa7 g3 a4 Ne7 Rxe7 Kg8 Bxd4+ Kf1 Bxb2
12  -85    298      796010     c6 < - > Rxa7 g3 a4 Ne7 Rxe7 Kg8 Rd7 Bxd4+ Rxd4 Kf8
12  -85    376      1066032    c6 Rxa7 g3 a4 Ne7 Rxe7 Kg8 Rd7 Bxd4+ Rxd4 Kf8
12  -84    418      1213919    a6 < + > Rxc7 g3 Kf1 a5 a4 Ne7 Rxe7 Rd8
12  -9     433      1261071    a6 < + > Rxc7 g3 Kf1 a5 a4 Ne7 Rxe7 Rd8
12  -3     476      1376898    a6 Rxc7 a5 Kf2 a4 Kg3 a3 bxa3 c3 d5 c2 Rxc2
13  -27    801      2240549    a6 < - > Rxc7 a5 a4 g3 Kf1 Ne7 Rxe7 Rd8 Rxd4
13  -102   1645     4526844    a6 < - > Rxc7 a5 a4 g3 Kf1 Ne7 Rxe7 Rd8 Rxd4 Nxg3
13  -102   1742     4884764    a6 Rxc7 a5 a4 g3 Kf1 Ne7 Rxe7 Rd8 Rxd4 Nxg3
13  -101   1850     5280251    c6 < + > Rxa7 g3 a4 Ne7 Rxe7 Kg8 Rd7 Bxd4+ Rxd4 Kf8
13  -101   2064     5859265    c6 Rxa7 g3 a4 Ne7 Rxe7 Kg8 Rd7 Bxd4+ Rxd4 Kf8
14  -125   3968     11235728   c6 < - > Rxa7 g3 a4 Ne7 Rxe7 Kg8 Rd7 Bxd4+ Rxd4 Kf8
14  -127   6017     18040839   c6 Rxa7 Ne7 Rxe7 Ra8 Nd6 Kg8 Nf5 Bf8 Re6 Rxa2 Rxc6 Rxb2 Rxc4
15  -137   12401    38007890   c6 Rxa7 Ne7 Rxe7 Ra8 Nd6 Kg8 Nf5 Bf8 Re6 Rxa2 Nxh6+ Bxh6 Rxh6 c3 Rg6+ Kf7
16  -161   25922    80872453   c6 < - > Kf1 c3 bxc3 g3 a4 Ne7 Rxe7 Rb8 Ke2 Rb2+ Kf3 Kh8 Nxg3 a6
16  -236   27022    83849163   c6 < - > Kf1 c3 bxc3 g3 a4 a6 a5 c5 dxc5 Ne7 Rxe7 Rb8 Nxg3 Kh8
16  -476   27308    84667641   c6 < - > Kf1 c5 dxc5 a6 Ke2 a5 c6 a4 c7 Ne7 Rxe7 Rd8 cxd8=Q a3 g3
16  -29988 31737    103290328  c6 <M6> Kf1 c5 dxc5 a6 a4 c3 bxc3 a5 g3 Ne7 Nf6+
Total nodes&#58;113917712 n/sec&#58;331400  &#40;q-nodes&#58;17% max depth&#58;31&#41;

Version same as 2.3 but with previous-version's more conservative pruning and faster zugzwangdetection.:

Code: Select all

Legal moves: 8   Static score: 236
Ply Score  Time     Nodes      Best move and expected line
------------------------------------------------------------
10  10     115      280271     a6 Rxc7 a5 Kf2 a4 Kg1 g3 Rxc4 a3 bxa3
11  10     189      514570     a6 Rxc7 a5 Kf2 a4 Kg1 g3 Rxc4 Ne7 Rxa4 Rd8
12  -3     398      1197782    a6 Rxc7 a5 Kf2 a4 Kg3 a3 bxa3 c3 d5 c2 Rxc2
13  -27    656      1933836    a6 < - > Rxc7 a5 Kf2 a4 a3 c3 bxc3 g3+ Nxg3 Nf6
13  -102   1456     4166111    a6 < - > Rxc7 a5 a4 g3 d5 c3 bxc3 Ne7 Rxe7 Rf8
13  -170   2957     8965786    a6 d5 a5 a4 g3 Rxc7 Ne7 Rxe7 Rb8 Rc7 c3 Nf6+ Kh8
13  -169   3811     12260476   c6 < + > Rxa7 c3 bxc3 g3 a3 Ne7 Rxe7 Rc8
13  -132   5932     18274677   c6 Rxa7 Ne7 Rxe7 Ra8 Nf6+ Kh8 Re6 g3 Rd6 Rf8 Ne4 Rf4
14  -156   8509     26227081   c6 < - > g3 a6 Kh1 c3 bxc3 c5 dxc5 a5 a4 Ne7 Rxe7 Rf8 Kg2 Kg8
14  -156   9684     30394983   c6 g3 a6 Kh1 c3 bxc3 c5 dxc5 a5 a4 Ne7 Nf6+ Bxf6+
15  -180   11911    37154831   c6 < - > g3 a6 Kh1 c3 bxc3 c5 dxc5 Ne7 Rxe7 Rf8 Rf2
15  -255   12922    40170870   c6 < - > g3 a6 Kh1 c3 bxc3 c5 dxc5 Ne7 Rxe7 Rf8 Ra7 a5 Kg2 a4 Rxa4
15  -29988 13172    40831611   c6 < - > g3 a6 Kh1 c3 bxc3 c5 dxc5 a5 a4 Ne7 Nf6+ Bxf6+
15  -29988 19889    69327674   c6 <M6> g3 a6 Kh1 c3 bxc3 c5 dxc5 a5 a4 Ne7 Nf6+
Total nodes&#58;76338784 n/sec&#58;353300  &#40;q-nodes&#58;12% max depth&#58;29&#41;

Stan

Hi everybody,

in the following game

[White "J. Hartweg"]
[Black "N.N."]
[Result "1-0"]
[PlyCount "51"]

1. e4 e5 2. Nc3 Nc6 3. f4 exf4 4. Nf3 g5 5. h4 g4 6. Ng5 h6
7. Nxf7 Kxf7 8. d4 d5 9. Bxf4 dxe4 10. Bc4 Kg6 11. O-O Bg7
12. h5 Kh7 13. Be5 Bxe5 14. Rf7 Bg7 15. Nxe4 Qxd4
16. Qxd4 Nxd4 17. Raf1 Bf5 18. c3 b5 19. cxd4 bxc4
20. R1xf5 Re8 21. Re5 Rf8 22. Ree7 Rxf7 23. Rxf7 1-0

black (a human) resigned after 23. Rxf7.

[D] 6nr/p1p2Rbk/7p/7P/2pPN1p1/8/PP4P1/6K1 b - - 7 23

On an AMD 64 X2 Dual Core Processor 5200+, 2.6GHz it took
glaurung2-epsilon/5 8:19 to see it is mated.

Code: Select all

  20	00:12	  30.912.639	2.523.480	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  21	00:18	  49.162.816	2.593.933	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  22	00:35	  93.432.205	2.639.999	-1,27	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  23	01:04	 172.684.871	2.668.225	-1,31	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  24	03:04	 491.877.670	2.670.983	-1,45	c7c6 Rf7xa7 Ng8e7 Ra7xe7
  25	08:19	1.334.990.761	2.671.680	-M6	c7c6 Kg1f2 a7a5 Kf2g3 a5a4 a2a3 c4c3 b2xc3 c6c5 d4xc5 Ng8f6 Ne4xf6+

How long does other programs need to see it's over?

By the way, how can I start a new thread in this forum?

kind regards
Bernhard