Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Support / Rybka 4 FAQ
1 2 3 4 5 Previous Next  
- - By Vasik Rajlich (Silver) [pl] Date 2010-05-26 07:12
Please ask questions here for a Rybka 4 FAQ.

Vas

ps. The next two days look packed for me, I'll probably do this the weekend of 28-30.5.
Parent - - By Caissafan (***) Date 2010-05-26 07:35
Well, starting now ...
1. there is a need for crystal-clear instructions how to enable 'large pages', which conditions, how to do
2. some clear words on the 'Asymmetric Piece Scoring' (APS) - which parameters were sensible for which situations
3. hints on 'Rook Endgame Scaling' parameter being aware of the new Rook Endgame heuristics
Parent - - By BigBen (****) Date 2010-05-26 09:19 Edited 2010-05-26 09:22

>1. there is a need for crystal-clear instructions how to enable 'large pages', which conditions, how to do


Hi,
    That is pretty simple and there are a few ways to go, one easy way is below

1. On the Start menu, click Run. In the Open box, type gpedit.msc.

The Group Policy dialog box opens.

2. On the Group Policy console, expand Computer Configuration, and then expand Windows Settings.

3. Expand Security Settings, and then expand Local Policies.

4. Select the User Rights Assignment folder.

The policies will be displayed in the details pane.

5. In the pane, double-click Lock pages in memory.

6. In the Local Security Policy Setting dialog box, click Add.

7. In the Select Users or Groups dialog box, add an account with privileges to run sqlservr.exe.

Regards
Parent - - By Robot Date 2010-05-26 09:53
what type of account to add
Parent - - By Vempele (Silver) [fi] Date 2010-05-26 10:08
The account you're going to use to run Rybka.
Parent - By Robot Date 2010-05-26 10:20
i am not able to understand it can u plzz elabraote a little
i just extracted and load the engine and it is not performing good
i am not able to understand its parameter settings and large page function
Parent - - By Wayne Lowrance (***) Date 2010-06-04 04:33
Vempele I do not understand. I have no idea what account I am running R4 in. How I check it, How I
determine  it/ set it ?.
Parent - By Uly (Gold) [mx] Date 2010-06-04 08:04
Just add "Everyone" and it should work.
Parent - By JoschiLup (*) [de] Date 2010-05-26 19:04
All this is not possible for Windows 7 Home, there is no Local Security Policy available!!!
Any workaround here?

Thanks
Jochen
Parent - By cma6 (****) Date 2010-05-30 02:56
Big Ben:
  This is excellent but is it only for Windows 7 or also for Win XP Pro 64-bit?
              Thanks, CMA
Parent - - By Banned for Life (Gold) Date 2010-05-30 03:15
The Windows policy Lock Pages in Memory option is disabled by default. This privilege must be enabled to configure Address Windowing Extensions (AWE). This policy determines which accounts can use a process to keep data in physical memory, preventing the system from paging the data to virtual memory on disk. On 32-bit operating systems, setting this privilege when not using AWE can significantly impair system performance. Locking pages in memory is not required on 64-bit operating systems.

The above is from the MSDN page on locking page in memory. Is there any reason that you don't accept the bolded sentence?
Parent - - By Bobby C (****) Date 2010-05-30 06:24

>Is there any reason that you don't accept the bolded sentence?


I've read this and don't understand. Is this bold sentence telling me that since I have Windows 7 Home Premium 64bit I don't have to lock pages to be able to use "large pages" on Rybka 4 ?

PS. I read before that you use a program that defrags memory. Can you give me a link to it? I use Game booster but apparently it's memory defrag is not good enough because I still can't get large pages working.
Parent - - By Uly (Gold) [mx] Date 2010-05-30 06:38
Have you tried using Large pages with a low RAM like 64MB?
Parent - By Bobby C (****) Date 2010-05-30 06:51
Nope and I also have not tried without giving some of my available memory to EGTB's. I will try disabling EGTB's and then I will try using large pages with something less then 1gb.
Parent - - By yanquis1972 (****) [us] Date 2010-05-30 07:50
do you have a program that defrags ram to the point that it's not necessary to reboot after fragmentation?
Parent - - By Uly (Gold) [mx] Date 2010-05-30 08:26
No, though I consider having R4 loaded at all times acceptable.
Parent - By yanquis1972 (****) [us] Date 2010-05-30 08:30
heh - i'm beginning to manage.
Parent - - By Vempele (Silver) [fi] Date 2010-05-30 08:29
Or better yet, 2MB?
Parent - By Uly (Gold) [mx] Date 2010-05-30 09:39
Oh yeah, for just seeing if they work at all 2MB would be optimal.
Parent - - By Banned for Life (Gold) Date 2010-05-30 14:45
If you believe the MS people know what they are talking about (granted, this is a big if), then you shouldn't need to lock pages on 64-bit systems. I suspect that locking pages on 64-bit systems is simply ignored. So it doesn't hurt, but why make the process more complicated than it needs to be?

I use 7-max, which is distributed under GPL, but it hasn't been updated in almost 5 years. It's available here: http://www.7-max.com/
It works fine with XP64, but I'm not confident it will work with later versions of windows.
Parent - - By Uly (Gold) [mx] Date 2010-05-30 22:38
Microsoft also says Large Pages doesn't work in XP...
Parent - By Banned for Life (Gold) Date 2010-05-31 00:47
It's an unsupported feature, but the important thing to remember is that huge pages is an Intel feature that is available for the memory manager, which is generally built into the OS. In my case, I have been using a third party memory manager which intercepts calls for dynamic memory allocation and returns pointers to huge pages. This doesn't require MS support.

What Vas has implemented in R4 uses MS's memory manager, and this seems to work in XP64 (don't know if it works in XP).
Parent - By cma6 (****) Date 2010-05-30 14:42
Since I never saw that post, I could not either accept or decline the bolded pages. Not everyone reads every single post on the forum.
My takeway is that one doesn't have to Lock Pages in Windows 7 or Windows XP Pro 64-bit.
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:47

> 1. there is a need for crystal-clear instructions how to enable 'large pages', which conditions, how to do


Some info: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=17033

More info this weekned.

> 2. some clear words on the 'Asymmetric Piece Scoring' (APS) - which parameters were sensible for which situations
> 3. hints on 'Rook Endgame Scaling' parameter being aware of the new Rook Endgame heuristics


More info this weekend.

Vas
Parent - - By Labyrinth (*****) [us] Date 2010-05-26 13:14
1. Did you still use C# extensively to work on R4?

2. Have you seen google's new programming language go, (3.) if so what do you think of it?

4. Did you finally break the 1000 mark on header files?

5. How did it feel when you sent in the final Rybka4 version to be sold?
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:48
1. Yes. I know it's boring but I kind of like C#.
2. No.
3. Not quite.
4. Pretty good but there wasn't much time to rest yet. :)

Vas
Parent - - By Bouddha (****) [ch] Date 2010-05-26 18:52
Hi.

Thanks.

How do we know if Large pages is setup & working correctly ?

regards
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:48
You can check the log files. There are other hints as well - we'll document this weekend.

Vas
Parent - - By Carl Bicknell (*****) [gb] Date 2010-05-26 18:59
Q: When do you estimate the release of Rental-Rybka?
Q: What improvements have been kept for Renal-Rybka that Rybka 4 UCI does not have?

Q: you stated ages ago that Rybka 4 was much tactically stronger than Rybka 3. In my tests it misses quite a lot of stuff Rybka 3 finds and although it is generally quicker in some positions its overall score was about the same as Rybka 3. Comments?
Parent - By Nelson Hernandez (Gold) [us] Date 2010-05-26 19:05
Renal-Rybka processes urine better than Rybka 4 UCI.
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:50
Rental Rybka: I'll have more to say about this soon.

Tactics: Yes, Rybka 4 should be clearly stronger tactically than Rybka 3. There will be some exceptions, and search has a lot of "luck" from one run to the next. If you have a specific position, I can check it out.

Vas
Parent - By Carl Bicknell (*****) [gb] Date 2010-05-28 09:05

> Tactics: Yes, Rybka 4 should be clearly stronger tactically than Rybka 3. There will be some exceptions, and search has a lot of "luck" from one run to the next. If you have a specific position, I can check it out


Well there are quite a few positions that Rybka 4 is either slower on or cannot solve which Rybka 3 can. Here are 2 for example:

1r3r1k/1bq4p/p4Pp1/4p3/1p1pN3/P3bQN1/1PP5/1K1R1B1R w - - 0 1


BM Bc4

and

r3r1k1/2qb1pp1/pp2n1np/2p1P3/2R1N3/1B3N1P/PP1Q1PP1/4R1K1 w - - 0 1


BM Nf6+
Parent - - By Nelson Hernandez (Gold) [us] Date 2010-05-26 19:03
Here are my questions:

1.  Why did you NOT implement bishop underpromotion AGAIN?  :)  :(

2.  How much ELO did you deliberately hold back from Rybka 4, and these withheld improvements relate to what specific areas?

3.  What will be the initial ELO-gap between Rybka 4 and Remote Rybka?

4.  Rybka 5: will it happen and if so, approximately when?

5.  Is it true that you have positively, absolutely ruled out any Rybka 4 updates?

6.  If a competing engine catches up to public-version Rybka 4 in strength before you release Rybka 5 how would you respond?

7.  Over the next year what are your development priorities with respect to Remote Rybka and Rybka 5?

8.  What do you consider to be Rybka 4's biggest remaining weakness, and do you plan to address this in Rybka 5?
Parent - - By Martin Thoresen (***) [se] Date 2010-05-26 19:04
Interesting questions, Nelson. Hoping for some answers from Vas. :)
Parent - By Milton (***) [us] Date 2010-05-26 19:11
+1
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:56

> 1.  Why did you NOT implement bishop underpromotion AGAIN?   


Sorry, it just wasn't a priority. :)

> 2.  How much ELO did you deliberately hold back from Rybka 4, and these withheld improvements relate to what specific areas?
> 3.  What will be the initial ELO-gap between Rybka 4 and Remote Rybka?


Elo is a question for testers.

> 4.  Rybka 5: will it happen and if so, approximately when?


The rough plan: fall of 2011. But please don't quote me on this, I am terrible with dates.

> 5.  Is it true that you have positively, absolutely ruled out any Rybka 4 updates?


No, but with the amount of testing that we did I would be shocked if something really big slipped past us.

> 6.  If a competing engine catches up to public-version Rybka 4 in strength before you release Rybka 5 how would you respond?


This probably wouldn't affect my plans or our releases.

> 7.  Over the next year what are your development priorities with respect to Remote Rybka and Rybka 5?


Remote Rybka is coming, for sure. We have now a concrete, realistic plan in place. More details will be ready soon.

Rybka 5 is coming as well, a bit further down the road.

> 8.  What do you consider to be Rybka 4's biggest remaining weakness, and do you plan to address this in Rybka 5?


It depends on how you define "biggest remaining weakness". Do you mean relative to top human players? This would be endgame play and strategic play. I'll address that. Or do you mean in terms of potential for further improvement? That would be normal search and normal eval. I'll address that too. Or do you mean as an analysis tool? I guess I'll address that too. :) All one small step at a time.

Vas
Parent - - By dcorbit (***) [us] Date 2010-05-27 18:25
On:

>> 1.  Why did you NOT implement bishop underpromotion AGAIN?   


>Sorry, it just wasn't a priority.


How about bishop underpromotion as an option for analysis.  When it is a contest, then the strongest playing is the goal.  When analyzing, the right answer is far more important.
Parent - - By vroger007 (**) [be] Date 2010-05-27 22:07
+1 !

Vas, will rybka5 have BU ?
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-06-04 13:23
Maybe. :)

Vas
Parent - - By Felix Kling (Gold) [de] Date 2010-06-04 13:26
...and I always thought it's TBD :)
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-06-04 13:37
That's too commital. :)

Vas
Parent - - By Carl Bicknell (*****) [gb] Date 2010-06-23 22:26
Hyperthreading. I get slightly higher nps with it ON under fritz GUI but I know R3 def prefers it off.
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-06-24 07:31
Can you post details?

Lukas has tested this quite carefully and HT always hurt slightly (when using one process per logical core).

Vas
Parent - By Geomusic (*****) Date 2010-06-04 19:21
R5 will serve you cocktails :)
Parent - - By George Tsavdaris (****) Date 2010-05-28 14:19
Did you XXXX forget about answering the 2) ?

XXXX=
a) intentionally
b) by mistake 

:-D

If we see the ELO chart of Rybka through time we can obviously see for the last 4th version, that you must have hit an iceberg in strength terms.
Did you? :-P
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-06-04 13:24
If I must have then I did. :)

Vas
Parent - - By Mark (****) [us] Date 2010-05-26 19:09
Are the Rybka 4 nps calculated the same way as in Rybka 3?  (i.e. Are they directly comparable?)
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-05-27 02:58
I didn't change the node counting but whenever the engine changes, you can't really compare NPS directly any more. The only thing it's good for is measuring your hardware.

Vas
Parent - By Mark (****) [us] Date 2010-05-27 03:10
Thanks!
Parent - - By Milton (***) [us] Date 2010-05-26 19:25
If I buy Aquarium Deep Rybka 4 for dvd, can I get an immediate download link so I won't have to wait for snail mail to start playing with the engine?

Thanks,
Milton
Up Topic Rybka Support & Discussion / Rybka Support / Rybka 4 FAQ
1 2 3 4 5 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill