Author: Zappa
Date: 06:33:11 02/17/06
Go up one level in this thread
On February 17, 2006 at 04:23:50, Vasik Rajlich wrote:
>On February 16, 2006 at 15:06:41, Zappa wrote:
>
>>On February 16, 2006 at 14:39:34, Vasik Rajlich wrote:
>>
>>>On February 16, 2006 at 13:17:10, Zappa wrote:
>>>
>>>>On February 16, 2006 at 06:21:57, Djordje Vidanovic wrote:
>>>>
>>>>>On February 15, 2006 at 10:20:18, Zappa wrote:
>>>>>
>>>>>>So I dropped by to see how my old buddies in CCC were doing, and found a massive
>>>>>>flamewar :) This is far more to my liking than dry test results, so I thought
>>>>>>I'd weigh in with a few comments.
>>>>>>
>>>>>>The majority of members in CCC seem to be unable to think rationally about
>>>>>>Rybka. I saw the following argument about 500X: "Vasik is violating the
>>>>>>charter" "But Rybka is so awesome". Clearly this is complete idiocy. The law
>>>>>>is intended to apply to everyone equally, regardless of whether or not they have
>>>>>>an awesome engine.
>>>>>>
>>>>>>I left (and this is not a coming back post) because I didn't like reading tons
>>>>>>of testing posts. However, the majority of people here appear to like it here.
>>>>>> From what I read in the flamewar Skinner seems correct that Vasik violated the
>>>>>>charter, but perhaps rather than crucifying Vasik (or Skinner), the charter
>>>>>>should be amended so the club serves the interest of the majority. It really
>>>>>>seems like the majority of people don't mind his "commerical posts" - so what's
>>>>>>the problem?
>>>>>>
>>>>>>I also read Enrico's post on Rybka's nodecount. Vasik claimed it was the result
>>>>>>of a bug. I read that as "my node count obfuscation algorithm has a bug and
>>>>>>will be fixed posthaste". If you look at Enrico's numbers, its obvious that
>>>>>>Rybka is really searching at 2-3M+ nps on big hardware.
>>>>>>
>>>>>>So, Vasik, please tell me: why would you would want to hide the node count of
>>>>>>Rybka?
>>>>>>
>>>>>>anthony
>>>>>
>>>>>
>>>>>Interesting post. And very telling. Demands some kind of reply, at least from
>>>>>a curious member of the CCC (but surely not from Vasik, as the question you ask
>>>>>is rather impertinent and irrelevant for the masses): why wouldn't you,
>>>>>Anthony, want to release your program so that Chrilly Donninger might try his
>>>>>disassembling skills on Zappa as he did with Rybka? A reminder for you: the
>>>>>verdict was that Rybka had no other program's code in it...
>>>>>
>>>>>If Rybka's code is messy and obfuscated, so be it. What's your code like remains
>>>>>to be seen, of course :-). LOL.
>>>>>
>>>>>Djordje
>>>>
>>>>I really don't understand your post at all. I am not saying that Rybka is a
>>>>clone. Nor am I saying that his source code is messy. I am just curious why he
>>>>deliberately modifies his node count/nps. No one else does this. Sure, some
>>>>people have different styles of counting. But no one does printf("nps / 20"),
>>>>and that appears to be what Vasik is doing.
>>>>
>>>>anthony
>>>
>>>It's not what I'm doing. I really don't know how you came to that conclusion.
>>>Although it's true that my node counting is a bit unusual, with a little bit of
>>>effort I could make it more conventional.
>>>
>>>Vas
>>
>>I didn't save Enrico's post, but Rybka's total nodes went something like this:
>>
>>50,000
>>100,000
>>150,000
>>200,000
>>10,000
>>12,000
>>14,000
>>
>>I would think that no matter how creative your counting scheme is, it should
>>still increase monotonically.
>>
>>anthony
>
>Ok, I understand your hypothesis. It's a bit more complicated than that. You can
>find a bunch of other examples on this forum (posts by Enrico Carrisco and Uri,
>IIRC) which will show this.
>
>Actually, if you go in a debugger, you can trivially see that two quantities are
>being combined. One I call "gulp", this is for me the interesting figure (for my
>private tests). The second is a simple ticker for the next I/O check.
>
>Vas
Gulp? LOL. That is some creative naming at least :)
I think I am just as confused as before. Why would you include a ticker for the
next IO check (time based) in the node count? Wouldn't that mean that Rybka's
nodes/depth are now no longer consistent across CPUs of different speeds? Seems
to me that would be a QA nightmare.
To be honest I am not as curious about the actual counting scheme - crazy though
it may be :) - as to why you would choose to use one.
anthony
P.S. I have never disassembled Rybka. I think I could, if I really wanted to,
but I have plenty of other things to do, and it sounds like it would be rather
boring and quite time-consuming.
This page took 0 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.