Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Aquarium / Cannot analyze with rybka since update.
- - By Drew (**) [ca] Date 2008-08-23 00:07
Hi there.Well thanks for fixing the database issue I had.Works great now!I have an hp laptop with dual amd 64 .I am running xp 32 bit windows.Since I upgraded to 3.04 with the downloader I am unable to analyze with rybka.As I step through the game often no data shows then invariably the whole thing stops and I get repeated pop ups.Access violation etc.I sent the first ones to you so you will have a record.When I check with task manager it says Aquarium is using 50% cpu and one thread of Rybka is using 50% cpu.The other thread is idle.I have tried resetting persistant hash,hash then reloading the engine.Nothing I know of helps.Can you point me in the right direction Please.
Parent - - By Drew (**) [ca] Date 2008-08-23 02:53
I found the problem I think.Seems Rybka uses the hash setting twice once for each thread.I have 2 gigs of ram but 256+32 for egtb put Rybka total pretty high and with other applications loaded into ram Aquarium did not have sufficient memory to do her thing.I lowered the hash setting to 128 and that did the trick.Thanks and have a good weekend.You lads deserve it!
Parent - - By Vempele (Silver) [fi] Date 2008-08-23 09:30

> Seems Rybka uses the hash setting twice once for each thread.


That's a common misconception. Even more so because it's been dropped from the FAQ. http://rybkachess.com/index.php?auswahl=FAQ+for+v+2.x

She does, however, use 75MB (+EGTB cache) per process in addition to the hash.
Parent - By Felix Kling (Gold) [de] Date 2008-08-23 09:33
you mean

"When I allocate a hash size of 512 MB, the task manager shows that each of the Rybka processes is using 512 MB. Is this correct?
No - this is a bug (or odd feature) in the task manager. There is only one hash table, which is shared between the Rybka processes.

In other words, you can allocate roughly half of your RAM for the hash table without overflowing your available RAM, despite the task manager's indications to the contrary."

I'll add this to the FAQ. Thanks for the tip :)
Parent - - By Drew (**) [ca] Date 2008-08-23 09:42
Hi there.Well in any event I am still getting access violations all over the place so I guess the ram thing made no difference.In addition to no output from rybka at various points in the game sometimes the buttons for navigating the game notation do not respond.Glaurung seems to work without problems.Any ideas?
Parent - - By cma6 (****) Date 2008-08-24 00:47
Did you do a Windows update on 8/15/08 on that computer?
I mention that because after my Windows update, I started having the same problem.
I had to do a Windows Restore and uninstall Deep Aquarium to get Rybka to work again. I still have not updated Windows.
Parent - - By Drew (**) [ca] Date 2008-08-24 12:11
Thank you. I will check that possibility out.Regards Drew
Parent - By Kapaun (****) [de] Date 2008-08-24 18:14
No, that can't be the reason. At least not generally. I'm always updating Windows, but I don't have a problem with IDea 3.04.
Parent - - By Drew (**) [ca] Date 2008-08-23 10:05
I should note also that the engine causing me grief is R3 960 w32.
Parent - - By buffos (Silver) [gr] Date 2008-08-23 10:16
i think this was an important detail
Parent - - By Drew (**) [ca] Date 2008-08-23 19:49 Edited 2008-08-23 20:09
Another observation.When I disable preserve analysis and Persistant hash, reset the values, then load the engine I do not get the access violation reports or the GUI unresponsiveness to the forward and backward buttons.However, different positions throughout the game will show no engine output.
Parent - - By buffos (Silver) [gr] Date 2008-08-23 20:19
i just saw your post in the Rybka thread.
You gave there another clue, about Databases... etc.

I think if you want to help developers find the problem , to tell all details.

I think there was a change in FRC pgn handling (fixed castling issues i think),  but if its related in any manner, it should be fixed

But its important to describe all the circumstances under which the problem happens (and when it does not happen)

Ofcourse developers will see it on Monday i guess, but a nice description would help a lot
Parent - By Drew (**) [ca] Date 2008-08-23 20:45
The database bug was fixed.I do not believe that it has any relation to the problem I am describing because you can copy a pgn and paste it and still the problem exists.I do agree that as much info as I can furnish is important but I do not have the luxury of spending all my time on this issue.As I see new things I report it.That's the best I can do.
Parent - - By Drew (**) [ca] Date 2008-08-23 22:58
Not sure what I did.This evening I decided to try and gather more info.Now everything works as it should.I just do not have any idea what's happened.I am running Aquarium just as I did before my posts but now Rybka 3 960 w32 works without the aforementioned problems.Sorry to waste everyone's time,my own included.All I can say is if anything strange happens going forward I will wait until the next evening before I post :-)
Parent - - By Dadi Jonsson (Silver) [is] Date 2008-08-23 23:34
:) If another user runs into something similar, then your posts about this issue might prove useful.
Parent - - By Kapaun (****) [de] Date 2008-08-24 11:12
Yes, we can advise him to give it a rest for a day or two, and then everything will be working fine...
Parent - - By Drew (**) [ca] Date 2008-08-25 22:08
Hi.Could you please analyze this Chess960 game with R3 960 w32. especially I am interested in your output for the position at move 26Qg5 then even more so with Rd7. Thanks

[Event "FiNet Open"]
[Site "Chess Classic Mainz 2008"]
[Date "2008.07.31"]
[Round "5"]
[White "Movsesian,S. (3.5)"]
[Black "Bareev,E. (3.5)"]
[Result "1/2-1/2"]
[FEN "qnbrkbnr/pppppppp/8/8/8/8/PPPPPPPP/QNBRKBNR w DHdh - 0 1"]
[SetUp "1"]

1. e4 b6 2. Nc3 Bb7 3. f3 e6 4. d4 d5 5. Bf4 Rd7 6. exd5 Bxd5 7. Bb5 Bc6 8. a4
Nf6 9. Nge2 Be7 10. O-O O-O 11. Be5 Rfd8 12. Nf4 Bxb5 13. axb5 c6 14. Ne4 Nd5
15. Nxd5 exd5 16. bxc6 Nxc6 17. Ng3 Qc8 18. Qc1 Nxe5 19. dxe5 Rc7 20. c3 d4 21.
Qf4 dxc3 22. bxc3 Rxc3 23. Nf5 Rc7 24. h4 Rxd1 25. Rxd1 Bf8 26. Qg5 Rd7 27. Kh2
Qe8 28. Re1 g6 29. h5 Qe6 30. Ng3 Be7 31. Qf4 Rd5 32. h6 a5 33. Qa4 Qd7 34. Qf4
a4 35. Ne4 Qf5 36. Qg3 Qh5+ 37. Kg1 Qxe5 38. Qf2 Qe6 39. Qb2 Qe5 40. Nf6+ Bxf6
41. Rxe5 Bxe5 42. Qc2 Bd4+ 43. Kf1 Bc5 44. Qxa4 Bf8 45. g4 b5 46. Qa8 Rc5 47.
Qb8 b4 48. Ke2 Rc2+ 49. Kd3 Rh2 50. g5 Rg2 51. f4 b3 52. f5 gxf5 53. Kd4 b2 54.
Ke5 f4 55. Kf4 Rf2+ 56. Ke4 Rc2 57. Kd4 Rg2 58. Ke3 Rg3+ 59. Kf4 Rg2 60. Ke3
Rxg5 61. Qxb2 Bxh6 62. Kf3 Rg6 63. Qe5 Bg7 64. Qe8+ Bf8 65. Kf4 h5 66. Qd8 Kg7
67. Qd4+ Kh7 68. Qd5 Bd6+ 69. Kf3 Rf6+ 70. Kg2 Kg6 71. Kh3 Bf4 72. Kh4 Rf5 73.
Qe4 Bg5+ 74. Kh3 Bf6 75. Qc2 Be5 76. Kh4 Kf6 77. Qc6+ Kg7 78. Qg2+ Kh7 79. Qe4
Kg6 80. Qg2+ Kh6 81. Qc6+ Bf6+ 82. Kh3 Rf4 83. Qd5 Rh4+ 84. Kg3 Rg4+ 85. Kh3
Rg5 86. Qxf7 1/2-1/2
Parent - - By Dadi Jonsson (Silver) [is] Date 2008-08-25 22:19 Edited 2008-08-25 22:21
That's interesting. 27. Nh6+ mate in one.

Edit: I forgot that I saw a -8 or -9 score before that move.
Parent - - By Drew (**) [ca] Date 2008-08-25 22:24
Thank you Dadi that at least confirms what I was seeing.I got that from both my troubled XP and Ubuntu with wine.I could not see anything wrong with the pgn can you?
Parent - - By Dadi Jonsson (Silver) [is] Date 2008-08-25 22:35
I had a quick look at the game up to move 27 and couldn't see anything wrong. And of course Aquarium is parsing it successfully.
Parent - By Drew (**) [ca] Date 2008-08-25 22:52 Edited 2008-08-25 23:18
Yes I have had no troubles so far with Aquarium and 960 Database handling.I think your team fixed that issue and I am pleased with the result. Perhaps after this issue is addressed in the next update or so it may resolve the other troubles I am still experiencing. Thanks for your help
Parent - - By buffos (Silver) [gr] Date 2008-08-25 22:26 Edited 2008-08-25 22:28
how did you enter the game? I try pasting the game score to sandbox and i cannot.
I have previously pasted the fen but

1) If i copy the whole game , it simply will not paste
2) If i paste the first moves (till Bf4 i think) the position is reverted to the normal chess position

[edit: if i past the whole pgn, then it pastes fine)
Parent - By Dadi Jonsson (Silver) [is] Date 2008-08-25 22:31
I converted the database from a PGN that Drew posted a few days ago. Now I tried to copy it from his post and paste it into the Sandbox and that works too. I'm attaching the game to this post.
Attachment: drew.pgn (1k)
Parent - - By ernest (****) [fr] Date 2008-08-23 18:55
> She does, however, use 75MB (+EGTB cache) per process in addition to the hash.

Shouldn't you say: She does, however, use 75MB per process in addition to the hash (+EGTB cache)?
(only one hash, one EGTB cache are used)
Parent - - By Vempele (Silver) [fi] Date 2008-08-23 20:23
I assumed the EGTB cache isn't shared (it'd require changing Nalimov's code). I don't have tablebases so I can't test it.
Parent - By ernest (****) [fr] Date 2008-08-24 12:28 Edited 2008-08-24 12:35
> I assumed the EGTB cache isn't shared

Sorry Teemu, you are RIGHT!!! :-p

(I just checked with the Task Manager)
Up Topic Rybka Support & Discussion / Aquarium / Cannot analyze with rybka since update.

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill